Adsorptive Separation of Light Olefin/Paraffin Mixture. Dispersion of CuCl in Faujasite Zeolites

Adsorptive Separationof

Light Olefin/Paraffin Mixtures

Dispersion of CuCl in Faujasite Zeolites

Adsorptive Separationof

Light Olefin/Paraffin Mixtures

Dispersion of CuCl in Faujasite Zeolites

PROEFSCHRIFT

ter verkrijging van de graad van doctor aan de Technische Universiteit Delft,

op gezag van de Rector Magnificus prof. dr. ir. J. T. Fokkema, voorzitter van het College voor Promoties,

in het openbaar te verdedigenop dinsdag 11 september om 12:30 uur

door

Arjen VAN MILTENBURG

scheikundig ingenieurgeboren te Utrecht

Dit proefschrift is goedgekeurd door de promotoren: Prof. dr. F. Kapteijn Prof. dr. J.A. Moulijn

Samenstelling promotiecommissie:

Rector Magnificus Technische Universiteit Delft, voorzitter Prof. dr. F. Kapteijn Technische Universiteit Delft, promotor Prof. dr. J.A. Moulijn Technische Universiteit Delft, promotor Prof. dr. J.C. Jansen Technische Universiteit Delft/Universiteit van Stellenbosch Prof. dr. ir. H. van Bekkum em. hgl. Technische Universiteit Delft Prof. dr. ir. G.V. Baron Vrije Universiteit Brussel Prof. dr. A.E. Rodrigues University of Porto Prof. dr. ir. P.J.A.M. Kerkhof Technische Universiteit Eindhoven

Prof. dr. W. Zhu has provided substantial guidance and support in the research presented in this thesis.

Dit onderzoek is uitgevoerd in de sectie Catalysis Engineering, DelftChemTech, Faculteit Technische Natuurwetenschappen, Technische Universiteit Delft met financiële steun van de Technische Universiteit Delft, de International Research Training Group (IRTG) “Diffusion in Porous Materials” (NWO-DFG), en het Network of Excellence “Inside Pores” (EU).

Proefschrift, Technische Universiteit Delft Met samenvatting in het Nederlands / with summary in Dutch ISBN: 978-90-6464-153-4

© 2007 Arjen van Miltenburg All rights reserved

Printed by Ponsen & Looijen B.V., Wageningen, the Netherlands

Contents

Chapter 1 Introduction 1

1.1 General introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1.2 Production of light olefins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1.3 Light olefin/paraffin separation processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 1.3.1 Absorption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 1.3.2 Physical adsorption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 1.3.3 Supported -complex adsorbents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 1.3.4 Membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 1.4 Multi-component adsorption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 1.5 Outline of the thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 1.6 List of symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 1.7 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

Chapter 2 Synthesis and characterization of CuCl/Faujasite for selective 17 olefin adsorption

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 2.1.1 Zeolites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 2.1.2 Cu+-ion dispersion in zeolites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 2.1.2.1 Ion-exchange . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 2.1.2.2 Monolayer dispersion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 2.2 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 2.2.1 Synthesis of zeolite NaX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 2.2.2 Synthesis of CuCl/Faujasite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 2.2.3 Adsorption of light olefins and paraffins studied with FTIR . . . . . . . . . . . . . . 23 2.2.4 Gases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 2.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 2.3.1 Synthesis of zeolite NaX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 2.3.2 Synthesis of CuCl/Faujasite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 2.3.3 Adsorption of light olefins and paraffins studied with FTIR . . . . . . . . . . . . . . 34 2.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 2.4.1 Synthesis of zeolite NaX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 2.4.2 Synthesis of CuCl/Faujasite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 2.4.3 Adsorption of light olefins and paraffins studied with FTIR . . . . . . . . . . . . . . 41 2.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 2.6 Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 2.7 List of symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 2.8 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

Chapter 3 Stability of CuCl/Faujasite adsorbents 45

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 3.2 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 463.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 3.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 3.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 3.6 Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 3.7 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

Chapter 4 Adsorption of olefins and paraffins on NaX and CuCl modified NaX 59

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 4.2 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 4.3 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 4.3.1 Adsorbents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 4.3.2 Volumetric method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 4.3.3 Adsorptives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 4.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 4.4.1 Isotherms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 4.4.2 Thermodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 4.4.3 Ideal adsorption selectivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70 4.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 4.5.1 Adsorption isotherms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 4.5.2 Thermodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 4.5.3 Ideal adsorption selectivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74 4.5.4 Mixture selectivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 4.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 4.7 Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 4.8 List of symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 4.9 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

Chapter 5 Binary adsorption of ethylene/ethane and propylene/propane 83 mixtures on NaX and CuCl/NaX

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 5.2 Mass balances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86 5.3 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 5.3.1 Breakthrough setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 5.3.2 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92 5.3.3 Breakthrough column fillings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 5.3.4 Breakthrough and desorption experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 5.3.5 Adsorptives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

5.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 5.4.1 Calculation procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 5.4.2 36 wt% CuCl/NaX columns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102 5.4.3 NaX-column . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 5.4.4 Single component breakthrough . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 5.4.5 Adsorption capacities and selectivities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110 5.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116 5.5.1 36 wt% CuCl/NaX adsorbent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117 5.5.2 NaX adsorbent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 5.5.3 Single component breakthrough . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120 5.5.4 Adsorption capacities and selectivities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120 5.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 5.7 List of symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124 5.7.1 Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124 5.7.2 Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 5.7.3 Flow sheet abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 5.8 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125

Chapter 6 Light olefin/paraffin separation: Summary, process options and 129 evaluation

6.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1306.2 Evaluation and industrial application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132 6.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137 6.4 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138

Samenvatting 139

List of publications and presentations 143

Dankwoord 147

Curriculum vitae 149

1Introduction

Introduction

1.1 General introduction Light olefins, like ethylene and propylene are important feedstocks in the chemical

industry. In the last decade worldwide production of ethylene and propylene has increased by ~50% to a capacity of more than 75 million tons of ethylene and over 47 million tons of propylene per annum (Table 1.1). In the coming future the production and demands are expected to continue to increase further at a rate of 3-5% per year (The Association of Petrochemical Producers in Europe (Appe) (2006)).

Table 1.1: 2004 worldwide production of light olefins. Geographical location Ethylene

[Mt a-1]Propylene [Mt a-1]

Asia 18 13Western Europe 21 15North America 32 17South America 4 2Total 75 47(The Association of Petrochemical Producers in Europe (Appe) (2006))

More than 50% of the monomeric olefins are used directly in polymers (e.g. HDPE, LDPE or PP). The remaining is first converted to other base chemicals like vinyl chloride, styrene, ethylene oxide, propylene oxide, cumene, acrylonitrile and alcohols (The Association of Petrochemical Producers in Europe (Appe) (2006)). These chemicals are then further processed into polymers or chemical products. The use of the light olefins in polymer production sets strict demands to the purities of the olefins. High purities (> 99 wt%) are required and the presence of acetylene should be limited, because it would deactivate the polymerization catalyst and because of safety aspects. Chemical grade olefins allow a lower purity (92-94 wt%) (Kroschwitz (1995)).

1.2 Production of light olefins In the petrochemical industry light olefins are obtained from a number of streams and

production units. Most of the light olefins are nowadays produced by steam cracking of hydrocarbons and by fluid catalytic cracking (FCC) of gas oils (Aitani (2004); Moulijn, Makkee, and van Diepen (2001)). A more recent technology is the dehydrogenation of the corresponding paraffin (Bhasin et al. (2001); Buyanov and Pakhomov (2001); Kotelnikov et al. (2001); Kotelnikov et al. (2004); Wang and Zhu (2004); Weyten et al. (2000)). For this several reaction schemes have been developed (Moulijn, Makkee, and van Diepen (2001)). The dehydrogenation can either be performed directly via a catalyst, or an oxidizing agent is used, which can react with the hydrogen atoms of the paraffin. Oxygen and carbon dioxide have been applied as oxidizing agent. Unfortunately, the dehydrogenation reaction is limited by the thermodynamic equilibrium. Single pass conversions of only 20-40% are obtained, separation of the product is still required and in the production plant large recycle streams are

2

Chapter 1

unavoidable. A forthcoming process to produce the olefins is the methanol to olefin (MTO) process.

QuenchWatersystem

Dilutionsteam

generation

Acid gasremoval

Drying

Acetylenehydrogenation

Fuel gas

EthaneButaneLPGNaphtha

Crackingfurnace Oil fractionation

Deethaniser56 trays2.5 MPa

Depropaniser55 trays2 MPa

Debutaniser

C3 Splitter180 trays2 MPa

Demethaniser & C2 Splitter120 trays

2 MPa

C5+

C4

Propane

Propylene

Ethane

Ethylene

Hydrogen, methane

C2-

C3+

C4+

C3 fraction

C2 fraction

243 K

Fuel oil

321 K

321 K260 K

QuenchWatersystem

Dilutionsteam

generation

Acid gasremoval

Drying

Acetylenehydrogenation

Fuel gas

EthaneButaneLPGNaphtha

Crackingfurnace Oil fractionation

Deethaniser56 trays2.5 MPa

Depropaniser55 trays2 MPa

Debutaniser

C3 Splitter180 trays2 MPa

Demethaniser & C2 Splitter120 trays

2 MPa

C5+

C4

Propane

Propylene

Ethane

Ethylene

Hydrogen, methane

C2-

C3+

C4+

C3 fraction

C2 fraction

243 K

Fuel oil

321 K

321 K260 K

Fig. 1.1: Process flow sheet of a typical steam cracking and olefin/paraffin separation plant (Eldridge (1993); Orica Limited (1999)).

3

Introduction

A typical process flow scheme for the production of ethylene and/or propylene via steam cracking is shown in Fig. 1.1. The feed of a cracker furnace consist of either naphtha, liquefied petroleum gas (LPG) or ethane and butane. In the cracker the larger molecules are cracked and dehydrogenated to methane, ethylene, propylene, C4-gases and other products. A naphtha feed would yield a diverse spectrum of products, while the gases result in larger yields of ethylene and propylene.

After cracking the heavy liquid components are removed first and the cracked gas is further cooled with quench water, followed by compression and scrubbing with caustic soda to remove the acidic gases (H2S and CO2). Before entering the deethaniser to split off the C3+-fraction, the gas is dried to remove water. After the deethaniser the highly undesired acetylene is removed from the shorter hydrocarbons (C2--fraction) by a selective hydrogenation. Acetylene would seriously deactivate the catalyst that is used in the production of polymers. The C2--fraction is then further cooled and distilled to remove hydrogen and methane. Finally ethylene is obtained in the C2-splitter from the remaining ethane/ethylene mixture. Ethane can be reused in the cracker as a recycle stream.

The C3+-fraction is fed to a depropaniser to obtain a propane/propylene mixture. In the C3-splitter propylene will be obtained. From the heavier gases a C4-fraction, containing C4-olefins, can be obtained in the debutanizer. Propane and these heavier gases can either be used for gasoline or other products or they can be recycled to the cracker.

Table 1.2: US energy demand for distillation.* (Total US energy demand ~ 100,000 PJ a-1.)Feed Typical components Estimated US

energy demand [PJ a-1]Petroleum Gasoline/naphtha 520

Crude Oil Light naphtha/heavy naphtha/light distillate

440

Liquefied petroleum gas (LPG) Ethane/propane/butane 230

OlefinsEthylene/ethane, propylene/propane

130

Miscellaneous hydrocarbons Cumene/phenol, acetone/acrylonitrile

110

Water-oxygenated hydrocarbons Methanol/water, water/acetic acid

110

Aromatics Ethylbenzene/styrene, benzene/toluene

80

Water-inorganics Ammonia/water 60Air Nitrogen/Oxygen 20Other 320 Total for distillation 2,000

* E.g. the first row gives the energy demand needed to separate a petroleum feed into a gasoline and naphtha fraction via distillation. (Eldridge, Siebert, and Robinson (2005); Energy Information Administration (EIA) (2006); Humphrey and Keller (1997)).

4

Chapter 1

Currently the separation of light olefin/paraffin mixtures is performed at high pressure in large and tall cryogenic distillation columns containing over one hundred trays (Fig. 1.1). The small divergence in relative volatility between ethylene and ethane ( < 1.5 (Barclay, Flebbe, and Manley (1982))) or propylene and propane ( = 1.09 (Sloley (2001))) makes it a very energy intensive process. Distillation is the largest energy consumer within the chemical industry. Besides the distillation of crude oil, petroleum and LPG (of which much larger amounts are produced for use as fuel), the US annual energy demand for the cryogenic olefin/paraffin separation is up to 130 PJ (Table 1.2). For The Netherlands this separation of ethylene/ethane and propylene/propane mixtures accounts for 5-15 PJ per annum (Vente, (2006)). Based on the production data provided in Table 1.1, and an annual growth of 3-5%, the annual worldwide energy demand for olefin/paraffin separation will be over 300 PJ for approximately 130 Mton of olefins.

Besides the high energy demands, the safety requirement for cryogenic distillation will also be very demanding, since the flammable compressed and cooled liquefied gas could explode in case of cooling failure, in particular since the olefin and paraffin themselves are integrated with the cooling system of the separation process.

1.3 Light olefin/paraffin separation processes The continuing rise in crude oil prices and the growing responsibility to reduce the

emissions of greenhouse gases resulted in many research projects to find alternative separation processes to fulfil the growing demand of olefins and to reduce the energy demands (Bryan (2004); Eldridge (1993)). Most attention is paid to absorption, adsorption and membrane based processes, often aided by a bonding of the olefin, e.g. via -complexation.

1.3.1 Absorption In absorption a selective interaction of the -orbitals of the double bond of the olefin with

cations such as Cu+ and Ag+ in solution, is formed. The selective absorption of olefins on the M+-ions results in the formation of a -complex (Herberhold (1974); Safarik and Eldridge (1998); Reine (2004)). A schematic picture of the complex is shown in Fig. 1.2. During the complexation of an olefin, the molecule will approach the surroundings of the M+-ion. At close proximity overlap of the molecular -orbitals of the olefin with the s- and d-orbital of the M+-ion will occur. The full -orbital of the olefin overlaps with the empty s-orbital of the M+-ion and an electron donation from the olefin towards the M+-ion occurs. At the same time the full dyz-orbitals of the M+-ion will overlap with the empty *-orbital of the olefin and a backdonation of electrons from this filled d-orbital towards the antibonding -orbital of the olefin will occur. The interaction of the olefin with the M+-ion will slightly loosen the CC-bond (Huang, Padin, and Yang (1999)), but it will remain intact. The -complex can be broken by simply lowering the pressure or raising the temperature.

In an olefin/paraffin separation process, the gas mixture will bubble through an extraction column filled with an absorption solution. In the column the olefins are absorbed, while the paraffin passes through. The solution is then pumped to another column, where the pressure is

5

Introduction

released and the olefin can desorb from the complex. The solution contains the M+-ion, either Cu+ or Ag+. To this purpose metal-salts, such as AgNO3, AgClO4 or AgBF4, are dissolved at a high concentration in water or another solvent. The use of Cu+ salts in water is not preferred, because of the low stability of Cu+ in water. Here options are CuTFA (cuprous trifluoroacetate) in propionitrile or aromatic solvents or CuAlCl4 in toluene. For AgNO3 ionic liquids have been reported to be attractive solvents (Kang et al. (2006); Won et al. (2005)).

C C

M

+

+ -

--

-+

+

+

-

+C C

M

+

+ -

--

-+

+

+

-

+

Fig. 1.2: Schematic picture of the Dewar-Chatt model for -complexation (Safarik and Eldridge (1998)).

1.3.2 Physical adsorption Separation via physical adsorption onto a porous solid can be based on a couple of physical

differences between the olefin and paraffin. For size exclusion only one component should be able to enter the internal pore structure of the adsorbent. If both components can enter the pore, differences in the diffusivity may allow kinetic separation. Alternatively, separation may be possible at equilibrium, either via an affinity or entropy driven adsorption.

For size exclusion a porous material which is only accessible for one of the components in an olefin/paraffin mixture is not commercially available yet and has to be synthesized. The other component should remain outside the adsorbent. So, an important criterion for such an adsorbent would be the size of the pore/window opening. It should be larger than the adsorbing component and smaller than the other components. An example of such an absorbent for propane/propylene separation is DD3R. In a previous study a nearly absolute separation of propane and propylene mixtures could be achieved (Zhu et al. (2000)). The 8-ring pore window of 0.36x0.44 nm only allows propylene to enter the zeolite cavities. The critical diameter of the methyl group of propane is larger than that of the methylene group of propylene and therefore the transport of propane through the window is severely hindered (Ter Horst et al. (2002)). Unfortunately, the diffusion of propylene in the zeolite is very slow. Furthermore the direct synthesis of DD3R could suffer from the formation of crystal polymorphs with DOH. Other examples of a microporous adsorbent for propane/propylene separation are AlPO-14 (Cheng and Wilson (1999)), and the zeolites 4A (Grande and Rodrigues (2005)), ITQ-3 and CHA (Olson et al. (2004)).

6

Chapter 1

Another possibility is to use the differences in diffusion rate of the molecules. When the pores of the adsorbent are close to the size of the molecules, the transport can be limited by the constraints of the pore size. Via this kinetic separation the fastest diffusing component will initially be the main component in the interior of the adsorbent, while the other component will primarily remain outside. In practical operation the adsorption should then be stopped before equilibrium is reached.

In reality there is a grey area of overlap between these two ways to separate olefin/paraffin mixtures. The influence of the temperature and the framework flexibility sometime allow the adsorption of a component whose critical diameter is larger than the rigid pore diameter mentioned in literature. However the diffusion of this component will be much slower and can be almost zero.

The last method to achieve a selective adsorption of the olefins is based on the affinity of and/or packing efficiency in the adsorbent towards the components in the mixture. Both components can penetrate the adsorbent at a sufficiently high diffusion rate and have an observable adsorption capacity. A separation based on the difference in adsorption affinity of the components relates to their adsorption enthalpies. In case of a separation based on the packing efficiency, the entropy is controlling the selectivity. The entropy effects become important at high loadings. In a practical process a binary equilibrium will be established in the adsorbent. Zeolites with a larger pore, which allow the adsorption of both components, like NaX or 5A (Da Silva and Rodrigues (1999); Huang et al. (1995); Järvelin and Fair (1993)), mesoporous structures (Grande et al. (2004); Newalkar et al. (2003)), and Kureha activated carbon (Zhu et al. (2005)) did show a selective adsorption of the olefins.

1.3.3 Supported -complex adsorbents The selectivity of adsorbents can be improved by the introduction of metal-ions that

selectively bind to the olefins. An example is the introduction of Cu+ or Ag+-ions on the support. They form a relatively stable -complex with the olefin, while the paraffin only adsorbs via the weak Van der Waals physical adsorption forces. To obtain a large number of adsorption sites on the supports, a high dispersion of the metal-ions is required. Examples of supports modified with M+-ions are: resins (Hirai, Kurima, and Komiyama (1986); Wu et al. (1997)), -Al2O3 (Blas, Vega, and Gubbins (1998)), SiO2 (Padin et al. (2000); Rege, Padin, and Yang (1998)), pillared clays (Choudary et al. (2002)), carbons (Hirai, Komiyama, and Keiichiro (1988); Mei et al. (2002)), mesoporous silica (Grande et al. (2005)) and zeolites (Pearce (1988); Takahashi, Yang, and Yang (2002)).

Adsorptive separation of other gas mixtures, like oxygen/nitrogen are currently primarily operated via Pressure Swing Adsorption (PSA) or Temperature Swing Adsorption (TSA) (Ruthven, Farooq, and Knaebel (1994); Thomas and Crittenden (1998)). Compared to the large and tall cryogenic distillation towers, more compact adsorption columns can be used. The adsorbent is installed as a fixed bed in a column and the adsorption is achieved at higher pressures or lower temperatures. The adsorbed component is afterwards recovered via a pressure decrease or temperature increase. Various design options involving multiple

7

Introduction

adsorption/desorption columns have been developed to obtain the highest purity and to use the compression/decompression power most efficiently. Adsorptive separation of light olefin/paraffin mixtures is still limited to the research stage. An adsorptive separation, especially if continuous membrane operation would be used, could result in a decrease of the energy cost to ~ 60% of that used nowadays for cryogenic distillation of light olefin/paraffin mixtures (Eldridge (2005)).

1.3.4 Membranes Membranes allow the conversion of a discontinuous adsorption based separation of light

olefin/paraffin mixtures into a continuous operation mode. In membranes the component with the larger diffusivity and/or strongest adsorption affinity will be selectively transported through the membrane layer, while the other components remain on the retentate side of the membrane. The continuous membrane operation has the advantage that no energy is lost in the compression/decompression cycles for PSA operations or the temperature cycles for TSA operations.

Olefin/paraffin selective membranes can be constructed from polymers (or carbonized polymers) or zeolites (Giannakopoulos and Nikolakis (2005)). Since the flow through the membrane is primarily controlled by the diffusion rate, a key factor for these membranes is to obtain a high selectivity while maintaining a high flux. Thicker membranes could result in a high selectivity, but they also increase the diffusion length. Thinner membranes could show a lower selectivity because of defects in the membrane layer. These defects are pinholes through which transport of all components occurs, which can become larger than the diffusive flow through the pores of the membrane thus reducing the selectivity of the membrane.

Polymeric membranes are also used to separate the gas stream from a solution containing a M+-salt (Duan, Ito, and Ohkawa (2003); Nymeijer et al. (2004)). The olefins will diffuse through the membrane and absorb in the solution. Via another membrane separation unit, or at the other side of a ‘supported liquid membrane’, the olefins will desorb from the solution and are collected at the low pressure side.

1.4 Multi-component adsorption In this thesis the main focus is on the separation of olefin/paraffin mixtures via selective

adsorption. For the separation of light olefin/paraffin mixtures via adsorption or via porous membranes, models are required to describe the mixture adsorption. For the adsorption of pure gases, multiple single component models have been proposed (Langmuir (Langmuir (1918)), Dual or multi-site Langmuir (Vlugt et al. (1998)), Toth (Toth (1971)), UNILAN (Honig and Reyerson (1952)), Virial (Barrer (1978)), etc. (Do (1998))) to describe the adsorption loadings at different pressures and temperatures. These models have been applied to all kind of gases adsorbing on various adsorbents. However in real separation processes more than one component will be present. Only when the contribution of the other component is constant or negligible the single adsorption of the desired component can be approximated with a single component isotherm model.

8

Chapter 1

In the other cases, like for the olefin/paraffin separation, information about the influence of the other components on the adsorption behaviour of each individual component will be required. The other components will also adsorb, and therefore the available volume (or surface area) for each component will be restricted by the others. Because of this competition between all components for the volume (or surface area) available in the adsorbent and the volume available in the gas phase, the adsorbed amounts will deviate from their single component adsorption isotherms.

To describe this multi-component adsorption several models have been developed (Yang (1987)). Starting from the single component Langmuir adsorption isotherms, two models have often been applied for binary adsorption. The single component Langmuir isotherms can be extended to a multi-component isotherm. The adsorption capacity of component i in a mixture of N components is then determined via the following equation:

N

jjj

iiii

pK

pKqq1

max (1.1)

The constants in the equation are then considered to be (a simple function of) the adsorption constants obtained from the measurement of single component isotherms. A similar modification can be applied to some of the other single component isotherm models.

This extension is only valid if the saturation capacities of both components are the same. If the saturation capacities differ, which is generally the case, entropy considerations enter the picture and only a thermodynamic consistent theory will be appropriate to predict the mixture adsorption. An alternative model for mixture adsorption that uses the data of the single component isotherms is the Ideal Adsorbed Solution (IAS) theory (Myers and Prausnitz (1965)). For this model the thermodynamic approach used for liquid-vapour equilibrium is extended to the gas-adsorbent equilibrium. For an ideal gas the chemical potential of each component in the adsorbed phase is given by:

)(ln)(),,( 001 iiigii PxTRTgxT (1.2)

where gi0(T) is the molar Gibbs free energy of component i at 101.325 kPa, i is the activity

coefficient of component i in the adsorbed phase, xi is the mol fraction of component i in the adsorbed phase and Pi

0( ) is defined as the equilibrium pressure for pure component i at the spreading pressure . In the ‘Ideal’ case the activity coefficient of all components is equal to one, while other values are used for the non-ideal or Real Adsorbed Solution (RAS).

9

Introduction

The spreading pressure of each component is calculated from the single component isotherm. Therefore the following integral has to be solved:

0

0

00 ln)()(iP

pii ppq

aRTP (1.3)

In this equation the single component isotherms are used to calculate loading qi0(p) at each

pressure. At the binary equilibrium pressure, the spreading pressure of all components will be equal. For the mixed gas phase the chemical potential is given by:

PyRTTgyT iii ln)(),,( 01 (1.4)

where yi is the mol fraction in the gas phase and P the total pressure. In case of thermodynamic equilibrium the chemical potential of the gas phase and the

adsorbed phase will be equal. For the Ideal Adsorbed Solution (IAS), the combination of Eq. 1.2 and 1.4 results in the following expression which should be satisfied for each component in the mixture:

)(0iii PxPy (1.5)

Further the following equations are required to calculate the Ideal Adsorbed Solution composition:

1ii xy (1.6)

01

i

i

t qx

q (1.7)

tii qxq (1.8)

where qi0 is the amount of component i adsorbed at spreading pressure in the absence of the

other components. To calculate the adsorption equilibrium of, for instance, a binary system an iteration

between Eq. 1.3, 1.5-1.8 and the pure component adsorption isotherms will be required. Eq. 1.8 finally yields the adsorbed amounts for the mixture adsorption at a given pressure, temperature and gas phase composition.

10

Chapter 1

1.5 Outline of the thesis In this introduction an overview of current practice and alternative techniques to separate

the light olefin/paraffin mixtures was given. As alternative, adsorptive separation is considered. Therefore the theoretical background of the Ideal Adsorbed Solution (IAS) theory for mixture adsorption was presented in this introduction chapter. The central theme of this thesis is the development, optimization and application of potential zeolite-based adsorbents for olefin/paraffin separation.

Specifically Faujasite zeolites and CuCl modified Faujasite zeolites are investigated for the separation of ethane/ethylene and propane/propylene mixtures. In Chapter 2 the synthesis optimization of uniform large zeolite NaX crystals is described. In these large crystals CuCl is dispersed and the adsorbent is characterized by different techniques such as X-Ray Diffraction (XRD), Thermo Gravimetric Analysis (TGA), DRIFT and FTIR transmission spectroscopy.

The observed colour changes and changes in the morphology of CuCl/NaX samples after prolonged storage in the atmosphere, resulted in a study to the stability of the CuCl/Faujasite zeolites. The effect of the exposure to humid air or water vapour on these adsorbents is presented in Chapter 3.

The single component adsorption results of ethane, ethylene, propane and propylene on NaX and the optimized CuCl/NaX adsorbent and their modelling are presented in Chapter 4.

For the actual application in an adsorption process, binary adsorption data will be required. Therefore a new adsorption breakthrough setup was constructed. In Chapter 5 a description of the setup is presented. Binary adsorption equilibrium data were determined for ethylene/ethane and propylene/propane (50:50) mixtures. The data obtained with these measurements are compared with the predictions of the Ideal Adsorbed Solution (IAS) theory using the single component isotherms presented in Chapter 4.

In Chapter 6 a summary of the previous chapters is given, the performance of the CuCl/NaX and NaX adsorbents for the separation of light olefin/paraffin mixtures is evaluated, possible process options are discussed, and the final conclusions are drawn.

1.6 List of symbols a Specific area of the adsorbent [m2 kg-1]gi

0 Molar Gibbs free energy of component i [J mol-1]Ki Adsorption constant for component i [Pa-1]N Number of components in gas mixture [-]P Total pressure [Pa-1]Pi

0 Equilibrium pressure for pure component i at [Pa-1]qi Adsorbed amount of component i [mol kg-1]qt Total adsorbed amount [mol kg-1]qi

sat Saturation capacity for component i [mol kg-1]Rg Universal gas constant [J mol-1 K-1]T Temperature [K]

11

Introduction

xi Mol fraction of component i in the adsorbed phase [-] yi Mol fraction of component i in the gas phase [-]

Relative volatility [-] i Activity coefficient of component i [-] i Chemical potential of component i [J mol -1]

Spreading pressure of the adsorbed phase [Pa m]

1.7 References Aitani, A. M., Advances in Propylene Production Routes, Oil Gas-Eur. Mag. 30 (2004) 36-39. Barclay, D. A., Flebbe, J. L. and Manley, D. B., Relative Volatilities of the Ethane-Ethylene

System from Total Pressure Measurements, J. Chem. Eng. Data 27 (1982) 135-142. Barrer, R. M., Zeolites and Clay Minerals as Sorbents and Molecular Sieves, Academic Press,

London (1978). Bhasin, M. M., McCain, J. H., Vora, B. V., Imai, T. and Pujadó, P. R., Dehydrogenation and

Oxydehydrogenation of Paraffins to Olefins, Appl. Catal. A-Gen. 221 (2001) 397-419. Blas, F. J., Vega, L. F. and Gubbins, K. E., Modeling New Adsorbents for Ethylene/Ethane

Separations by Adsorption via Pi-Complexation, Fluid Phase Equilibr. 150 (1998) 117-124.

Bryan, P. F., Removal of Propylene from Fuel-Grade Propane, Separ. Purif. Rev. 33 (2004) 157-182.

Buyanov, R. A. and Pakhomov, N. A., Catalysts and Processes for Paraffin and Olefin Dehydrogenation, Kinet. Catal. 42 (2001) 64-75.

Cheng, L. S. and Wilson, S. T., Vacuum Swing Adsorption Process for Separating Propylene from Propane, US Patent 6 296 688 (1999).

Choudary, N. V., Kumar, P., Bhat, T. S. G., Cho, S. H. and Han, S. S., Adsorption of Light Hydrocarbon Gases on Alkene-Selective Adsorbent, Ind. Eng. Chem. Res. 41 (2002) 2728-2734.

Da Silva, F. A. and Rodrigues, A. E., Adsorption Equilibria and Kinetics for Propylene and Propane over 13X and 4A Zeolite Pellets, Ind. Eng. Chem. Res. 38 (1999) 2051-2057.

Do, D. D., Adsorption Analysis: Equilibria and Kinetics, Imperial College Press, London (1998).

Duan, S., Ito, A. and Ohkawa, A., Separation of Propylene/Propane Mixture by a Supported Liquid Membrane Containing Triethylene Glycol and a Silver Salt, J. Membrane Sci. 215 (2003) 53-60.

Eldridge, R. B., Olefin/Paraffin Separation Technology: A Review, Ind. Eng. Chem. Res. 32 (1993) 2208-2212.

Eldridge, R. B., Brainstorming Session Background Information as Part of the Hybrid Technology Workshop Separations Research Program in Austin, TX, USA, (2005).

Eldridge, R. B., Siebert, F. A., and Robinson, S., Hybrid Separations/Distillation Technology, Research Opportunities for Energy and Emissions Reduction, (2005).

Energy Information Administration (EIA), Short-Term Energy Outlook, www.eia.doe.gov (2006).

12

Chapter 1

Giannakopoulos, I. G. and Nikolakis, V., Separation of Propylene/Propane Mixtures using Faujasite-Type Zeolite Membranes, Ind. Eng. Chem. Res. 44 (2005) 226-230.

Grande, C. A., Araujo, J. D. P., Cavenati, S., Firpo, N., Basaldella, E. and Rodrigues, A. E., New Pi-Complexation Adsorbents for Propane-Propylene Separation, Langmuir 20 (2004) 5291-5297.

Grande, C. A., Firpo, N., Basaldella, E. and Rodrigues, A. E., Propane/Propylene Separation by SBA-15 and Pi-Complexated AG-SBA-15, Adsorption 11 (2005) 775-780.

Grande, C. A. and Rodrigues, A. E., Propane/Propylene Separation by Pressure Swing Adsorption using Zeolite 4A, Ind. Eng. Chem. Res. 44 (2005) 8815-8829.

Herberhold, M., Metal Pi-Complexes: Part II: Specific Aspects, Elsevier, New York (1974). Hirai, H., Komiyama, M. and Keiichiro, W., Solid Adsorbent for Unsaturated Hydrocarbon

and Process for Separation of Unsaturated Hydrocarbon from Gas Mixture, US Patent 4 747 855 (1988).

Hirai, H., Kurima, K. and Komiyama, M., Selected Solid Ethylene Adsorption Composed of Copper (I) Chloride and Polystyrene Resin Having Amino Groups, Polym. Mater. Sci. Eng. 55 (1986) 464-468.

Honig, J. M. and Reyerson, L. H., Adsorption of Nitrogen, Oxygen, and Argon on Rutile at Low Temperatures; Applicability of the Concept of Surface Heterogeneity, J. Phys. Chem. 56 (1952) 140-146.

Huang, H. Y., Padin, J. and Yang, R. T., Comparison of Pi-Complexations of Ethylene and Carbon Monoxide With Cu+ and Ag+, Ind. Eng. Chem. Res. 38 (1999) 2720-2725.

Huang, Y.-H., Liapis, A. I., Xu, Y., Crosser, O. K. and Johnson, J. W., Binary Adsorption and Desorption Rates of Propylene-Propane Mixtures on 13X Molecular Sieves, Separ. Technol. 5 (1995) 1-11.

Humphrey, J. L. and Keller, G. E., Separation Process Technology, McGraw-Hill, New York (1997).

Järvelin, H. and Fair, J. R., Adsorptive Separation of Propylene-Propane Mixtures, Ind. Eng. Chem. Res. 32 (1993) 2201-2207.

Kang, S. W., Char, K., Kim, J. H., Kim, C. K. and Kang, Y. S., Control of Ionic Interaction in Silver Salt-Polymer Complexes with Ionic Liquids: Implications for Facilitated Olefin Transport, Chem. Mater. 18 (2006) 1789-1794.

Kotelnikov, G. R., Komarov, S. M., Bespalov, V. P., Sanfilippo, D. and Miracca, I., Application of FBD Processes for C-3-C-4 Olefins Production from Light Paraffins, Stud. Surf. Sci. Catal. 147 (2004) 67-72.

Kotelnikov, G. R., Komarov, S. M., Titov, V. I. and Bespalov, V. P., New Propylene Production Process, Petro. Chem. 41 (2001) 422-427.

Kroschwitz, J. I., Kirk-Othmer Encyclopedia of Chemical Technology, Wiley, New York (1995) Vol. 9 pp. 877 and Vol. 20 pp. 257.

Langmuir, I., The Adsorption of Gases on Plane Surfaces of Glass, Mica and Platinum, J. Am. Chem. Soc. 40 (1918) 1361-1403.

13

Introduction

Mei, H., Hu, C. G., Liu, X. Q. and Yao, H. Q., Study of Activated Carbon Supported CuCl for Ethylene/Ethane Separation by Adsorption: Effects of Oxidative Treatment, New Carbon Mat. 17 (2002) 33-37.

Moulijn, J. A., Makkee, M. and van Diepen, A., Chemical Process Technology, John Wiley & Sons, Chichester, England (2001).

Myers, A. L. and Prausnitz, J. M., Thermodynamics of Mixed-Gas Adsorption, AIChE J. 11 (1965) 121-127.

Newalkar, B. L., Choudary, N. V., Turaga, U. T., Vijayalakshmi, R. P., Kumar, P., Komarneni, S. and Bhat, T. S. G., Potential Adsorbent for Light Hydrocarbon Separation: Role of SBA-15 Framework Porosity, Chem. Mater. 15 (2003) 1474-1479.

Nymeijer, D. C., Visser, T., Assen, R. and Wessling, M., Composite Hollow Fiber Gas-Liquid Membrane Contractors for Olefin/Paraffin Separation, Sep. Purif. Technol. 37 (2004) 209-220.

Olson, D. H., Camblor, M. A., Vilaescusa, L. A. and Kuehl, G. H., Light Hydrocarbon Sorption Properties of Pure Silica Si-CHA and ITQ-3 and High Silica ZSM-58, Micropor. Mesopor. Mat. 67 (2004) 27-33.

Orica Limited, Ethylene Chemical Fact Sheet: Ethylene Production: Process Flow Diagram, (1999).

Padin, J., Rege, S. U., Yang, R. T. and Cheng, L. S., Molecular Sieve Sorbents for Kinetic Separation of Propane/Propylene, Chem. Eng. Sci. 55 (2000) 4525-4535.

Pearce, G. K., Selective Adsorption and Recovery of Organic Gases using Ion-Exchanged Faujasite, US Patent 4 717 398 (1988).

Rege, S. U., Padin, J. and Yang, R. T., Olefin/Paraffin Separation by Adsorption: Pi-Complexation vs. Kinetic Separation, AIChE J. 44 (1998) 799-809.

Reine, T.A., Olefin/Paraffin Separation by Reactive Adsorption, PhD Thesis, The University of Texas, Austin, USA (2004).

Ruthven, D. M., Farooq, S. and Knaebel, K. S., Pressure Swing Adsorption, VCH Publishers, New York (1994).

Safarik, D. J. and Eldridge, R. B., Olefin/Paraffin Separations by Reactive Absorption: A Review, Ind. Eng. Chem. Res. 37 (1998) 2571-2581.

Sloley, A. W., Distillation Operations Manual, www.distillationgroup.com (2001). Takahashi, A., Yang, F. H. and Yang, R. T., New Sorbents for Desulfurization by Pi-

Complexation: Thiophene/Benzene Adsorption, Ind. Eng. Chem. Res. 41 (2002) 2487-2496.

Ter Horst, J. H., Bromley, S. T., Van Rosmalen, G. M. and Jansen, J. C., Molecular Modelling of the Transport Behaviour of C-3 and C-4 Gases through the Zeolite DD3R, Micropor. Mesopor. Mat. 53 (2002) 45-57.

The Association of Petrochemical Producers in Europe (Appe), Western European Market Review, www.petrochemistry.net (2006).

Thomas, W. J. and Crittenden, B., Adsorption Technology and Design, Butterworth-Heinemann, Oxford (1998).

14

Chapter 1

Toth, J., State Equations of Solid-Gas Interface Layers, Acta. Chim. Hung. 69 (1971) 311-328.

Vente, J., Hydrocarbon Separation, Energy research Centre of the Netherlands (ECN), (2006). Vlugt, T. J. H., Zhu, W., Kapteijn, F., Moulijn, J. A., Smit, B. and Krishna, R., Adsorption of

Linear and Branched Alkanes in the Silicalite-1, J. Am. Chem. Soc. 120 (1998) 5599-5600. Wang, S. and Zhu, Z. H., Catalytic Conversion of Alkanes to Olefins by Carbon Dioxide

Oxidative Dehydrogenation-A Review, Energ. Fuel. 18 (2004) 1126-1139. Weyten, H., Luyten, J., Keizer, K., Willems, L. and Leysen, R., Membrane Performance: the

Key Issues for Dehydrogenation Reactions in a Catalytic Membrane Reactor, Catal. Today 56 (2000) 3-11.

Won, J., Kim, D. B., Kang, Y. S., Choi, D. K., Kim, H. S., Kim, C. K. and Kim, C. K., An Ab Initio Study of Ionic Liquid Silver Complexes as Carriers in Facilitated Olefin Transport Membranes, J. Membrane Sci. 260 (2005) 37-44.

Wu, Z. B., Han, S. S., Cho, S. H., Kim, J. N., Chue, K. T. and Yang, R. T., Modification of Resin-Type Adsorbents for Ethane/Ethylene Separation, Ind. Eng. Chem. Res. 36 (1997) 2749-2756.

Yang, R. T., Gas Separation by Adsorption Processes, Imperial College Press, London (1987).

Zhu, W., Groen, J. C., Van Miltenburg, A., Kapteijn, F. and Moulijn, J. A., Comparison of Adsorption Behaviour of Light Alkanes and Alkenes on Kureha Activated Carbon, Carbon 43 (2005) 1416-1423.

Zhu, W., Kapteijn, F., Moulijn, J. A., Den Exter, M. C. and Jansen, J. C., Shape Selectivity in Adsorption on the All-Silica DD3R, Langmuir 16 (2000) 3322-3329.

15

Introduction

16

Part of this research has been published earlier in: - Van Miltenburg, A., Zhu, W., Kapteijn, F., Moulijn, J.A., Stud. Surf. Sci. Cat. 158B (2005), 979-986. - Van Miltenburg, A., Zhu, W., Kapteijn, F., Moulijn, J.A., Chem. Eng. Res. Des. 84 (2006), 350-354.

2Synthesis and characterization of

CuCl/Faujasite for selective olefin adsorption

The synthesis of large NaX crystals and the dispersion of CuCl in Faujasite zeolites have been optimized. TGA and XRD show a maximum dispersion capacity of 36 wt% for NaX and 43 wt% for NaY, corresponding to approximately 10 CuCl molecules per super cavity. SEM and TEM analysis showed that the CuCl was well dispersed inside the zeolite crystal. A DRIFT study, with CO as a probe molecule confirmed that at 623 K CuCl was dispersed in the zeolite. It also showed that the adsorption of CO on CuCl is stronger than on the Na+-ionof the Faujasite zeolite, due to its strong -complexation with Cu+.

The selective adsorption of ethylene and propylene on the CuCl/Faujasite zeolites was confirmed in a low-pressure IR transmission cell. The paraffins could be quickly removed by reducing the pressure, while lower pressures were required to desorb the olefins.

Synthesis and characterization of CuCl/Faujasite for selective olefin adsorption

2.1 Introduction The separation of olefin/paraffin mixtures can be achieved by adsorption using the

selective interaction of the olefin with metal-ions, like Cu+ or Ag+, forming a -complex (Herberhold (1974); Chapter 1 of this thesis). The Cu+ is the cheaper element of these two and shows a higher affinity with the olefins (Yang (2003)) and is therefore chosen as the selective metal-ion in this study. In order to create a large number of adsorption sites for the olefin, these ions have to be dispersed over a large surface area. Supports with a large surface areas on which these metal-ions were dispersed earlier are resins (Hirai, Kurima, and Komiyama (1986); Wu et al. (1997)), -Al2O3 (Blas, Vega, and Gubbins (1998); Yang and Kikkinides (1995)), carbons (Hirai, Komiyama, and Keiichiro (1988); Mei et al. (2002)), SiO2 (Padin et al. (2000); Rege, Padin, and Yang (1998)), pillared clays (Choudary et al. (2002)), mesoporous silica (Grande et al. (2005)) and zeolites (Pearce (1988); Takahashi, Yang, and Yang (2002)). Compared to the other supports, zeolites have a very uniform accessible nanoporous crystal structure, large surface areas and a good thermal and chemical stability.

2.1.1 Zeolites There are several types of zeolites discovered so far, both natural and artificial (Baerlocher

and McCusker (2006)). Zeolites are widely used in the petrochemical industry as catalysts or adsorbents for several processes (Moulijn, Makkee, and van Diepen (2001)). For the use of the selective adsorption of olefins, the zeolite should have pores which are large enough to allow the adsorbing gases to enter the cavities via its window openings and to allow a fast diffusion of these gases in the interior of the zeolite crystal. The selectivity of the zeolites towards the olefins can be further improved by the dispersion of metal-ions in the zeolite cavities. A good interaction of these dispersed metal-ions with the support would be preferred in order to prevent the continuous loss of metal-ion adsorption sites, e.g. via sintering. Because of the positive charge of the metal-ion, a negatively charged structure would result in a good interaction. In zeolite structures negative charges can be formed in the structure by replacing part of the Si-atoms with Al-atoms. This is quantified by the so-called Si/Al-ratio of the zeolite. During the synthesis this negative charge will be compensated by positive ions (e.g. Na+) in the synthesis solution. These ions will remain in the cavities of the zeolite at the end of the synthesis, but can later on be exchanged by other ions.

Of the zeolite structures, the large 12-ring window and 3D pore structure of Faujasite (FAU) would allow a fast diffusion. A low Si/Al-ratio is expected to result in a good interaction of the metal-ion with the support. Furthermore the zeolite is commercially available and also multiple recipes can be found (Robson and Lillerud (2001)). A schematic picture of the Faujasite structure is shown in Fig. 2.1. The super cavity is formed by the combined stacking of sodalite cages and hexagonal prisms (D6R). The Si and Al atoms occupy the vertexes of these polyhedra. The O-atoms are located in the middle of each edge, between the Si- and/or Al-atoms. The sodalite cage can be entered by small gases via the 6-ring window (S6R) shared with the super cavity. For the Faujasite structure two versions can be found, namely zeolite X and zeolite Y. The difference is based on the Si/Al ratio of the

18

Chapter 2

structure. For zeolite X this ratio should be between 1.0-1.5, while for zeolite Y a ratio above 1.5 is used (Breck (1974)). The cations, which counterbalance the charge of the zeolite framework, are located in the sites indicated in Fig. 2.1. For zeolite Y the sites I, I’ and II are (partially) occupied (Fitch, Jobic, and Renouprez (1986)), while for zeolite X sites I, I’, II and III’ are (partially) occupied (Olson (1995)).

A fast adsorption equilibrium throughout the crystal is best achieved when small crystals are used. However small crystals would result in a severe pressure drop when they are used in packed bed adsorption columns, which is not preferred in research applications. To overcome this pressure drop, larger pellets can be pressed from the small crystals. In many cases a binder would be required to maintain the pellet structure. These pellets could result in an additional transport barrier for the adsorption and desorption of the gases and would complicate modelling. Another option is to synthesize large zeolite crystals. For the research application of the large zeolite crystals in adsorption measurements important factors are the phase purity and particle size distribution of the synthesized crystals.

S6RD6R

sodalitecage

supercavity

S6RD6R

sodalitecage

supercavity

Fig. 2.1: Schematic picture of the Faujasite crystal structure and the cation sites.

2.1.2 Cu+-ion dispersion in zeolites To improve the selectivity of the zeolite for the selective adsorption of olefins, Cu+-ions

will be dispersed in the zeolite structure. Several techniques have been developed, including ion-exchange and monolayer dispersion (Yang (2003)).

2.1.2.1 Ion-exchange Al-containing zeolites contain metal counter-ions to compensate the charge unbalance of

the zeolite structure. These ions can be freely exchanged by other ions once a driving force exists. The degree of the ion-exchange that can be obtained will depend on: the concentration

19


of the ions in the solution, the nature of the cations, the temperature, the anions of the salt, the solvent, and the structure of the zeolite. The exchange properties of zeolites are widely used in processes, both in industry (catalyst manufacture) and by consumers (softening of water).

A typical procedure for the ion-exchange of a zeolite would start with dissolving a salt containing the preferred metal-ion in water or an other solvent. The solution is contacted with the zeolite, either batch-wise by suspending the zeolite particles in the solution, or continuously by circulating the solution through a bed of zeolite crystals. After the exchange the zeolites are filtered and washed with water. In order to enhance the degree of ion-exchange in a batch (or semi-batch) process, the procedure is often carried out multiple times.

Unfortunately this procedure does not work for Cu+ salts, since they are insoluble in water and will oxidize to Cu2+ in the solution. A possible alternative is to perform the ion-exchange with a Cu2+-salt. The Cu2+-ions are exchanged as [Cu2+OH]+ and can be converted to Cu+ by thermal auto-reduction at 573-723 K in helium, whereby the following reactions occur (Takahashi et al. (2001)):

2 [CuOH]+ [Cu2O]2+ + H2O [Cu2O]2+ 2 Cu+ + ½ O2

After completion of this auto-reduction reaction the Cu+ exchanged form of the zeolite will be obtained.

2.1.2.2 Monolayer dispersion Dispersion of salt onto solid substances can be achieved using two techniques, either via

thermal dispersion (Chen and Sachtler (1998); Xie et al. (1996)) or via incipient wetness impregnation. For thermal dispersion of a salt onto a solid surface, the two solids are thoroughly mixed. The mixed solids are heated to a temperature between the Tamman temperature of the salt and its melting point (Tm). Above the Tamman temperature, approximately ½ Tm, the crystal structure of the salt becomes more flexible and mobile. The salt molecules can diffuse over the surface of the support and a dispersed layer is formed. This process is also known as a ‘solid state ion-exchange’.

For incipient wetness impregnation the salt is dissolved into a solvent. The support is wetted with the solution, and the solvent containing the salt fills the pores of the support by capillary forces. Thereafter the sample is slowly heated to remove the solvent by evaporation. (Part of) the salt will remain in the pores of the solid and will be deposited on the internal surface. The solvent should be chosen in such a way that the salts can dissolve in the solvent and allow sufficient wetting of the internal surface of the porous solid. To deposit Cu+ onto the surface a cuprous-ammonium-salt can be dissolved in water or a CuAlCl4/toluene solution can be used. Also impregnation with a CuCl2-solution followed by the thermal auto-reduction to CuCl in an inert atmosphere can be performed.

20

Chapter 2

In this study large crystal CuCl/NaX adsorbents will be synthesized. Therefore first the synthesis of large crystals of NaX will be optimized and these will be characterized using XRD, SEM and N2-physisorption. Thereafter CuCl will be dispersed on the surface of the pores of Faujasite zeolites using the thermal dispersion method. Compared to the other techniques this is the most simple and effective method and high loadings can be obtained. In most cases an ion-exchange or an incipient wetness impregnation would require an additional reduction step, which would complicate the synthesis procedure and would affect the reproducibility. Since the thermal dispersion process of CuCl on Faujasite is poorly characterized in the literature, it is further optimized using various characterization techniques, like Thermogravimetric analysis (TGA), XRD, SEM/EDX, TEM and FTIR spectroscopy. For the IR study, CO will be used as a probe molecule in the DRIFT-cell.

In order to characterize the adsorbent for its application in an adsorptive olefin/paraffin separation process, the adsorbent should show a higher affinity for one of the components in the mixture. A quick screening can be performed by studying the surface complexes on the adsorbent by means of FTIR spectroscopy. Unfortunately the gas phase of the olefins and paraffins also absorb part of the infrared light. A preliminary study in the DRIFT cell indicated that the absorbance of the adsorbed components could hardly be distinguished from the gas phase contribution at atmospheric pressures. Therefore the adsorption of ethane, ethylene, propane and propylene was tested in the low-pressure IR transmission cell.

2.2 Experimental

2.2.1 Synthesis of zeolite NaX Large zeolite NaX crystals were synthesized using two recipes reported in the literature

(Charnell (1971); Qiu et al. (1998)) and a modified version of the recipe of Qiu (Van Miltenburg et al. (2006)). Following the recipe of Qiu, an Al-solution was made by dissolving NaAlO2 (Riedel-de Haën) in a sodium hydroxide solution in water. Triethanolamine (J.T. Baker) was added as a stabilizing and buffering or complexing agent. The solution was filtered twice through a 0.2 m filter to remove remaining particulates and other impurities. A Si-solution was made by dissolving Aerosil 90 (kindly provided by Degussa) in water. Finally triethanolamine was added to the Si-solution. Both solutions were aged for about 1 hour after which they were mixed together. The molar ratio of the ingredients was: SiO2 : NaAlO2 : NaOH : TEA : H2O = 1 : 1.4 : 4.3 : 2 : 222. The gel was kept at 353 K so the crystallization of the zeolites could occur. After 2-3 weeks the crystals were filtered, washed thoroughly and dried at 353 K for 1 day. To remove small silica particles attached to the crystals, they were ultrasonically cleaned in ethanol and dried afterwards at 373 K in air. The obtained crystals were divided into several sieve fractions.

For the recipe of Charnell sodium silicate (Na2SiO3·9H2O, Acros) instead of Aerosil 90 was used as silica source. In the modified version of the recipe of Qiu two silica sources were used. Besides Aerosil 90 a few milligrams of Na2SiO3.9H2O were added to 250 ml of the Si-sol (Na2SiO3·9H2O : SiO2 ~ 1 : 5000).

21


The synthesized NaX crystals were analyzed with XRD (Bruker-AXS D5005, CuK ) and SEM (Philips XL20, 15kV). The Si/Al-ratio of the zeolite was determined by ICP-OES (PerkinElmer Optima 3000DV). The Na content in the zeolite was determined by AAS. The porous properties were determined by volumetric N2-physisorption at 77 K (Quantachrome Autosorb-6B).

The particle size distributions were determined with a Mastersizer S, with a 300 mm RF lens in a flow-through cell. The particles were dispersed in demineralised water. To break up agglomerated particles, the dispersion was exposed for 4 minutes to ultrasonic vibrations.

2.2.2 Synthesis of CuCl/Faujasite Physical mixtures were prepared by mixing different amounts of CuCl (Fluka) with the

synthesized NaX crystals or commercial NaY (Zeolyst CV100, Si/Al = 2.55). These physical mixtures were slowly heated (1 K min-1) in the quartz reactor to 623 K in flowing argon with a rate of 100 ml (STP) min-1 and at this temperature the samples were heated for typically 4 h. Thereafter heating was stopped and the temperature slowly returned to room temperature.

To limit the influence of the exposure to the ambient atmosphere to the minimum, some mixtures were also prepared in a ¼” stainless steel tube, which could be closed from the atmosphere with two valves. The physical mixtures were fixed in a 3 cm long tube between two 1 mm thick stainless steel frits, with pore openings of 0.5 m. After the dispersion procedure, the two valves were closed and the samples were removed and stored in a glove box under nitrogen to prevent the exposure of the CuCl to ambient air.

The TGA experiments of Faujasite and of the physical mixtures of CuCl and Faujasite were performed in a Mettler Toledo TGA/SDTA851e. Depending on the composition for the experiments, a sample amount ranging from 15 to 40 mg was inserted in an alumina TGA cup of 70 l. For all TGA experiments the volume of the sample in the cup was approximately equal; therefore the amount of Faujasite was similar for all experiments.

Once the samples containing NaY were inserted in the TGA, they were purged for 30 min at 298 K with a helium flowrate of 100 ml (STP) min-1. The temperature was then slowly raised (1 K min-1) to 373 K and this temperature was kept for 1 h. Then the temperature was further increased to 623 K at 2 K min-1. After that a temperature of 623 K was kept for typically 4 h. To investigate the effect of a longer dispersion time, samples with an excess of CuCl were heated at 623 K for 8h.

For NaX, and physical mixtures of NaX with CuCl, a slightly different procedure was used to gain a more accurate value of the water content of the zeolite at different temperatures. Once the samples containing NaX were inserted in the TGA, they were purged for 3 h at 298 K with a helium flowrate of 100 ml (STP) min-1. The temperature was then slowly raised (1 K min-1) to 423 K and this temperature was kept for 1.5 h. Then the temperature was further increased to 523 K at 1 K min-1 and this temperature was kept for 1.5 h. Then the temperature was further increased to the final temperature of 623 K at 1 K min-1 and kept there for 5 h. Preliminary tests indicated that the slightly different temperature program for NaX did not significantly affect the final mass decrease obtained for the various samples.

22

Chapter 2

The XRD patterns were recorded for CuCl/Faujasite and for physical mixtures of CuCl and Faujasite with a Philips PW1830/40 diffractometer using CuK radiation. The dispersed samples were synthesized beforehand in the quartz reactor, after which the XRD patterns were recorded immediately to limit exposure to the ambient air.

The DRIFT experiments of CO adsorption were performed in a Nicolet Nexus FTIR at 323 K. The DRIFT-cell was equipped with KBr windows and absorption spectra were recorded with a nitrogen cooled MCT detector. CO adsorption was performed on NaY and 43 wt% CuCl/NaY (weight percentage based on the dry NaY mass). To minimize the exposure to the air, the dispersion of CuCl on NaY was achieved inside the DRIFT cell by increasing the temperature to typically 623 K at 1 K min-1 in flowing helium with a rate of 100 ml (STP) min-1 and at this temperature the samples were heated for 4 h. Before measurement the samples were rapidly cooled (>150 K min-1) to 323 K, at which temperature the adsorption of a mixture of 5 vol% CO in helium on the adsorbent was monitored. The desorption was achieved by returning to the pure helium flow. Evolved gas analysis by a mass spectrometer confirmed that the gas phase composition change was established within seconds. In order to calculate the absorbance, the background spectra of the samples after the dispersion of CuCl on NaY, but before CO adsorption were recorded at 323 K. To investigate the effect of the CuCl-dispersion a physical mixture of 43 wt% CuCl and NaY was only heated to 423 K for 4 h to evaporate most of the adsorbed water, after which the temperature was rapidly decreased to 323 K for CO adsorption.

The synthesized CuCl/NaX was further characterized using SEM (Philips XL20, 15kV) and TEM (Philips CM30T, 300 kV, with LINK EDX system). For the TEM analysis a suspension of ground small crystals in hexane was dropped on a microgrid carbon polymer supported on an aluminium grid, followed by drying at ambient conditions, all in an argon glovebox. Samples were transferred to the microscope in a special vacuum-transfer sample holder under exclusion of air.

2.2.3 Adsorption of light olefins and paraffins studied with FTIR The adsorption behaviour of light olefins and paraffins was investigated using a low-

pressure IR transmission cell equipped with KBr windows (Miura and Gonzalez (1982)). An electric heating coil around the transmission cell allowed the control of the temperature inside the vacuum chamber up to 398 K. The transmission cell was installed inside an FTIR spectrometer (Nicolet Magna-IR 860) and absorption spectra were recorded using the DTGS detector. The transmission cell was connected to a gas introduction/evacuation system, of which the flowsheet is shown in Fig. 2.2. From a fixed volume (B) a small amount of a gas mixture could be introduced to the transmission cell.

In this study a thin wafer was pressed (0.5 GPa) from the NaY zeolite crystals with and without 43 wt% dispersed CuCl. Since the thin wafers pressed from the large synthesized NaX crystals were unstable, the small commercial NaY (Zeolyst CV100, Si/Al = 2.55) crystals were used. The wafer was placed in the transmission cell (see Fig. 2.2 point A) and the transmission cell was closed with the KBr windows. A low pressure in the sample cell was obtained by slowly opening the valves NV6, V5, V4 and V2 to remove the gas from the

23


lines of the setup and the transmission cell via the 1st rotary pump. When all valves were opened completely and the pressure slowly started to decrease, the temperature of the transmission cell was slowly increased to 398 K in steps of 10 K. During heating the pressure in the cell was continuously monitored so the heating could be stopped wherever the pressure started to increase too rapidly. At the higher temperatures the adsorbed water started to desorb and evaporated from the adsorbent wafer. Once the pressure in the mixture chamber (PB in Fig. 2.2) was below 10-2 mbar, valve V1 was slowly opened. This way the turbo pump could further reduce the pressure. Because of the slow pressure decrease and slow temperature increase applied in this experimental procedure, the pressure throughout the setup and across the wafer would be approximately equal and would therefore reduce the chance of cracking the thin wafer.

Once the pressure in the fixed volume (PB) remained stable at the minimum detection limit (1.0 10-7 mbar), the temperature was lowered to 323 K. In order to calculate the absorbance, a background IR spectrum was recorded at this temperature. Ethane, ethylene, propane and propylene adsorption were investigated on both samples. Before the introduction of a small amount of gas inside the fixed volume, the valves to the 1st rotary pump and the turbo pump were closed (V1, V2 and V5). By slowly opening valve V3, a small gas volume was introduced in the fixed volume (B). The pressure (PB) of the volume was increased to approximately 50 mbar. This gas sample was then introduced to the transmission cell by opening valve V5. Once a stable pressure was reached in the transmission cell, an IR absorption spectra was recorded. Thereafter the pressure was reduced by shortly opening

Fig. 2.2: Flowsheet vacuum infrared setup.

24

Chapter 2

valve V2. At several intermediate pressures, this valve was closed and an IR spectrum was recorded after the pressure had stabilized. When the pressure in the chambers was below 10-2

mbar, valve V1 was also opened completely and all gases were fully desorbed. The introduction of 50 mbar of gas and the slow reduction in pressure was performed and analyzed twice for each gas on each adsorbent to investigate the reproducibility of the experiment and the reversibility of the adsorption/desorption process. After the two duplicate experiments for each gas, the sample was reheated to 398 K to completely desorb all adsorbed components, after which another gas was investigated.

2.2.4 Gases The gases used in the experiments were all supplied by HoekLoos and had the following

purities: helium 4.6 (> 99.996%), 5 vol% CO 2.0 (> 99.0%) in helium 4.6 (> 99.996%), argon 4.6 (> 99.996%), ethane 3.0 (>99.9%), ethylene 2.8 (99.8%), propane 3.5 (>99.95%) and propylene 3.5 (99.95%).

2.3 Results

2.3.1 Synthesis of zeolite NaX Elemental analysis of all the synthesized zeolites indicated a Si/Al ratio of 1.2-1.3 and a

Na/Al ratio of 1.0 (Table 2.1), which is within the range for NaX. The porous properties (SBET

and Vmicro) are close to those obtained for commercial Faujasite zeolites. Only the zeolite synthesis based on the Charnell recipe shows a larger deviation from the expected values. The XRD patterns of the synthesized zeolites are shown in Fig. 2.3a-e. The larger crystals, obtained by the recipe of Qiu and by the modified recipe, were divided in a sieve fraction of crystals with a diameter smaller and larger than 50 m. The XRD-patterns of both sieve fractions are shown in the figures. For comparison the literature XRD patterns of the zeolites NaX (Faujasite), NaP (Gismondine) and NaA (Linde Type A) are included as well in Fig. 2.3f-h (Baerlocher and McCusker (2006)). The visible reflection peaks corresponding to NaP and NaA are marked by P and A, respectively, in the XRD patterns of the synthesized zeolites. The other reflection peaks correspond to the literature pattern of NaX. The SEM images of the synthesized zeolite crystals are pictured in Fig. 2.4a-e. As seen in the SEM pictures, the zeolites synthesized based on the Charnell recipe resulted in small (agglomerated) crystals of 2-3 m, while for the other recipes crystal sizes up to 100-120 mwere obtained in the synthesis. Of the two recipes with large crystals the SEM pictures of both sieve fractions are shown.

25


Table 2.1: Elemental analysis and porous properties of the zeolites. Geographical location Si/Al-ratio

[-] Na/Al-ratio

[-] SBET

[m2 g-1]Vmicro

[cm3 g-1]NaX (Charnell) 1.2 1.0 749 0.29 NaX (Qiu) 1.3 1.0 852 0.33 NaX (Modified) 1.3 1.0 857 0.33 NaY (Zeolyst, CV100) 2.55 1.0 875 0.33

a

0 10 20 30 40 50 600

20

40

60

80

100

Rel

ativ

ein

tens

ity[-]

0

20

40

60

80

100

Rel

ativ

ein

tens

ity[-]

2 [-]

a

0 10 20 30 40 502 [-]

60

b

0 10 20 30 40 50 62 [-]

00

20

40

60

80

100

Rel

ativ

ein

tens

ity[-]

PA P P P P

b

0 10 20 30 40 50 62 [-]

00

20

40

60

80

100

Rel

ativ

ein

tens

ity[-]

b

0 10 20 30 40 50 62 [-]

(Charnell)

00

20

40

60

80

100

Rel

ativ

ein

tens

ity[-]

PA P P P PPA P P P P

(Qiu, dp< 50 m)

c

0 10 20 30 40 50 602 [-]

0

20

40

60

80

100

Rel

ativ

ein

tens

ity[-]

PA P P P P

c

0 10 20 30 40 50 62 [-]

00

20

40

60

80

100

Rel

ativ

ein

tens

ity[-]

c

0 10 20 30 40 50 62 [-]

00

20

40

60

80

100

Rel

ativ

ein

tens

ity[-]

d

0 10 20 30 40 50 62 [-]

00

20

40

60

80

100

Rel

ativ

ein

tens

ity[-]

PA P P P P

d

0 10 20 30 40 50 62 [-]

00

20

40

60

80

100

Rel

ativ

ein

tens

ity[-]

d

0 10 20 30 40 50 62 [-]


(Qiu, dp> 50 m)

00

20

40

60

80

100

Rel

ativ

ein

tens

ity[-]


(Modified, dp< 50 m)

0 10 20 30 40 50 602 [-]

0

20

40

60

80

100

Rel

ativ

ein

tens

ity[-]

e

0 10 20 30 40 502 [-]

0

20

40

60

80

100

Rel

ativ

ein

tens

ity[-]

e

60

(Modified, dp< 50 m)

00 10 20 30 40 50 62 [-]

0

20

40

60

80

100

Rel

ativ

ein

tens

ity[-]

f

0 10 20 30 40 50 62 [-]

0

20

40

60

80

100

Rel

ativ

ein

tens

ity[-]

0

f(Literature NaX)

Fig. 2.3: XRD pattern of: (a) NaX (Charnell recipe), (b) NaX dp<50 m (Qiu recipe), (c) NaX dp>50 m (Qiu recipe), (d) NaX dp<50 m (modified recipe), (e) NaX dp>50 m (modified recipe) and (f) NaX (literature).

26

Chapter 2

0 10 20 30 40 50 602 [-]

0

20

40

60

80

100

Rel

ativ

ein

tens

ity[-]

g

0 10 20 30 40 50 62 [-]

0

20

40

60

80

100

Rel

ativ

ein

tens

ity[-]

0

g(Literature NaP)

00 10 20 30 40 50 62 [-]

0

20

40

60

80

100

Rel

ativ

ein

tens

ity[-]

h

0 10 20 30 40 50 62 [-]

0

20

40

60

80

100

Rel

ativ

ein

tens

ity[-]

0

h(Literature NaA)

Fig. 2.3: XRD pattern of (g) NaP (literature) and (h) NaA (literature) (Baerlocher and McCusker (2006)).

a

bb c

d e

Fig. 2.4: SEM images of: (a) NaX (Charnell recipe), (b) NaX dp<50 m (Qiu recipe), (c) NaX dp>50 m (Qiu recipe), (d) NaX dp<50 m (modified recipe) and (e) NaX dp>50 m (modified recipe).

27


The particle size distribution of the crystals obtained from the recipe of Qiu and our modified crystals are plotted in Fig. 2.5. Also the distribution after 4 minutes exposure to ultrasonic vibrations is included. Due to this ultrasonic treatment the average particle size decreases and additional fines were formed. The modified recipe resulted in a smaller fraction of fines before and after ultrasonic vibrations and gave a more uniform particle size distribution.

01

dp [μm]

Pro

babi

lity

[-]

10 100 1000

5

10

15

ab

d

c0

1dp [μm]

Pro

babi

lity

[-]

10 100 1000

5

10

15

ab

d

c

Fig. 2.5: Particle size distribution of: (a) Qiu recipe (short dashed line, ), (b) Qiu recipe 4 min ultrasonic treated (dotted line, ), (c) modified recipe (solid line, ) and (d) modified recipe 4 min ultrasonic treated (long dashed line, ).

2.3.2 Synthesis of CuCl/Faujasite The TGA patterns of NaY and of the physical mixtures of CuCl and NaY are shown in Fig.

2.6a. The hydrated NaY powders lost 21 wt% of the initial mass of the zeolite sample upon heating from 298 to 373 K, corresponding to regions I-III in Fig. 2.6a. A further temperature increase from 373 K to 623 K resulted in an extra mass loss of 3 wt%. For all the physical mixtures of CuCl and NaY, the mass loss below 623 K is about 24 wt% on the basis of the initial mass of NaY in the mixtures. For an amount of CuCl in the mixture below 43 wt%, the mass of the mixture sample remains constant at 623 K for 4 h, while for a higher amount of CuCl in the mixture, a decrease in the mass is still observed.

The TGA patterns of NaX (Charnell recipe) and of the physical mixtures of CuCl and NaX are shown in Figure 2.6b. The hydrated NaX powders lost 20 wt% of the initial mass of the zeolite sample upon heating from 298 to 423 K, corresponding to regions I-III in Figure 2.6b.A further temperature increase from 423 K to 623 K resulted in an extra mass loss of 3 wt% (regions IV-VI). For all the physical mixtures of CuCl and NaX, the mass loss below 623 K is about 23 wt% on the basis of the initial mass of NaX in the mixtures. For an amount of CuCl in the mixture below 36 wt%, the mass of the mixture sample remains constant at 623 K for 5 h, while for a higher amount of CuCl in the mixture, a decrease in the mass is still observed.

28

Chapter 2

In Fig. 2.7 the loading of a dispersion of CuCl in NaY is calculated for a dispersion time of 4 and 8 h. Below a loading of 43 wt%, the longer dispersion time has no effect on the loading of CuCl remaining on the surface of the dispersed CuCl/NaY zeolite. Above 43 wt%, a longer dispersion time results in a lower amount of CuCl still present in the mixture, due to the sublimation of the excess amount of CuCl.

Temperature [K

]

0 100 200 300 400 5000.7

0.8

0.9

1.0

Time [min]

Rel

ativ

e w

eigh

t[-]

300

400

500

600

I II III IV V

47 wt%

43 wt%

0 wt%

57 wt%

25 wt%

aNaY Tem

perature [K]

0 100 200 300 400 5000.7

0.8

0.9

1.0

Time [min]

Rel

ativ

e w

eigh

t[-]

300

400

500

600

I II III IV V

47 wt%

43 wt%

0 wt%

57 wt%

25 wt%

a

0 100 200 300 400 5000.7

0.8

0.9

1.0

Time [min]

Rel

ativ

e w

eigh

t[-]

300

400

500

600

I II III IV V

47 wt%

43 wt%

0 wt%

57 wt%

25 wt%

aNaY

Rel

ativ

ew

eigh

t[-]

00.7

Time [min]

300

Temperature

[K]

200 400 600 800 1000

0.8

0.9

1.0

400

500

600

I II III IV V VI VII

36 wt%

24 wt%

0 wt%

46 wt%

12 wt%

bNaX

Rel

ativ

ew

eigh

t[-]

00.7

Time [min]

300

Temperature

[K]

200 400 600 800 1000

0.8

0.9

1.0

400

500

600

I II III IV V VI VII

36 wt%

24 wt%

0 wt%

46 wt%

12 wt%

bNaX

Fig. 2.6: Thermogravimetric analysis profiles of heat treatment in helium of: (a) NaY (Zeolyst) and physical mixtures of CuCl and NaY, and (b) NaX (Charnell recipe) and of physical mixtures of CuCl and NaX.

00

Initital loading of CuCl [wt%]

Rem

aini

nglo

adin

gof

CuC

l[w

t%]

20 40 60

20

40

60

43wt% 8h

4h

00

Initital loading of CuCl [wt%]

Rem

aini

nglo

adin

gof

CuC

l[w

t%]

20 40 60

20

40

60

43wt% 8h

4h

Fig. 2.7: Remaining CuCl loading after heat treatment procedure in helium of physical mixtures of CuCl and NaY.

Fig. 2.8a represents the XRD patterns of the physical mixtures of CuCl and NaY with different amounts of CuCl. Both characteristic patterns of CuCl and NaY appear in the XRD patterns. The increased dilution of NaY with CuCl in the physical mixtures results in a decrease in the intensities of the reflections of NaY. Fig. 2.8b shows the XRD patterns of these mixtures after the heat treatment at 623 K for 4 h. Compared to the physical mixtures the intensities of the reflection of CuCl (2 = 28.5 and 47.4) have decreased considerably.

29


02 [-]

Inte

nsity

[A.U

.]

10 20 30 40 50

28.5 47.46.2 10.2

0 wt%

25 wt%

13 wt%

36 wt%

47 wt%

57 wt%a

23.7

02 [-]

Inte

nsity

[A.U

.]

10 20 30 40 50

28.5 47.46.2 10.2

0 wt%

25 wt%

13 wt%

36 wt%

47 wt%

57 wt%a

23.7 28.5

02 [-]

Inte

nsity

[A.U

.]

10 20 30 40 50

47.46.2 10.2

0 wt%

25 wt%

13 wt%

36 wt%

47 wt%

57 wt%b

43 wt%

41 wt%

23.7 28.5

02 [-]

Inte

nsity

[A.U

.]

10 20 30 40 50

47.46.2 10.2

0 wt%

25 wt%

13 wt%

36 wt%

47 wt%

57 wt%b

43 wt%

41 wt%

23.7

Fig. 2.8: XRD patterns of NaY (Zeolyst) with different amounts of CuCl: (a) physical mixtures, (b) CuCl dispersed in NaY.

Initial weight fraction of CuCl [-]

Inte

nsity

at 2

= 6.

2[·1

03A

.U.]

Physicalmixture

Dispersedsample

0 0.2 0.4 0.60

2.5

7.5

5.0

7.5 a


Inte

nsity

at 2

= 6.

2[·1

03A

.U.]

Physicalmixture

Dispersedsample

0 0.2 0.4 0.60

2.5

7.5

5.0

7.5 a


Inte

nsity

at 2

= 23

.7[·1

03A

.U.]

0 0.2 0.4 0.60

1

b

2

4

5

3


Inte

nsity

at 2

= 23

.7[·1

03A

.U.]

0 0.2 0.4 0.60

1

b

2

4

5

3

0Initial weight fraction of CuCl [-]

Inte

nsity

2=

28.5

[·103

A.U

.]

Inte

nsity

2=

47.4

[·10

3A

.U.]

0.2 0.4 0.60

5

10

15

0

2..5

5.0

7.5

Physicalmixture

Dispersedsample

c

0Initial weight fraction of CuCl [-]

Inte

nsity

2=

28.5

[·103

A.U

.]

Inte

nsity

2=

47.4

[·10

3A

.U.]

0.2 0.4 0.60

5

10

15

0

2..5

5.0

7.5

Physicalmixture

Dispersedsample

c

Fig. 2.9: Intensity of the NaY reflections at (a) 2 = 6.2 and (b) 2 = 23.7 for the physical mixture ( ) and the dispersed sample ( ) at different starting loadings of CuCl (weight fraction). (c) Intensity of the CuCl reflections at 2 = 28.5 ( , ) and 2 = 47.4 ( , ) for the physical mixture ( , ) and the dispersed sample ( , ) of CuCl and NaY. The lines are to guide the eye.

30

Chapter 2

The intensities of most of the reflections of NaY are the same for the physical mixtures and the heat treated samples. However the reflections at 2 = 6.2 and 10.2 are further reduced in the heat treated samples. In Fig. 2.9a-b the intensities of the reflections of NaY at 2 = 6.2 and 2 = 23.7 are shown for the physical mixture before and after the CuCl dispersion at 623 K. The reflections at 2 = 23.7 decrease similarly for both samples, so the NaY zeolite is still present in a similar ratio with CuCl as in the physical mixture. The low angle reflection at 2= 6.2 shows a further reduction in the intensity after the CuCl dispersion at 623 K and becomes almost zero at high CuCl loadings.

The intensities of the reflections of CuCl (2 = 28.5 and 47.4) are plotted in Fig. 2.9cversus the loading of CuCl for the physical mixtures before and after the heat treatment at 623 K for 4 h. For the physical mixtures a linear increase in the intensity of the reflections of CuCl is observed. For all the heated samples there is a decrease in the reflections of CuCl. Below a loading of 43 wt% CuCl the reflections of CuCl disappear almost completely. When the amount of CuCl in the mixture exceeds 43 wt%, the reflections of crystalline CuCl do not disappear completely and a relatively high intensity remains.

For the NaX (Charnell recipe) zeolite the XRD pattern of the dispersed mixtures with CuCl are shown in Fig. 2.10. Like NaY the reflections of CuCl disappear below a saturation loading of 36 wt% for NaX. Above this loading the reflection peaks of CuCl remain visible. The dispersion of CuCl into NaX does not affect the intensities of the reflections of NaX, since the decrease of the NaX reflections follows the same trend as observed for the physical mixture of NaY. The low angle reflections at 2 = 6.2 and 10.2 decrease more quickly than the other reflections of NaX for higher CuCl loadings.

02 [-]

Inte

nsity

[A.U

.]

10 20 30 40 50

46 wt%

36 wt%

24 wt%

12 wt%

0 wt%

28.5 47.46.2 10.2 23.7

02 [-]

Inte

nsity

[A.U

.]

10 20 30 40 50

46 wt%

36 wt%

24 wt%

12 wt%

0 wt%

28.5 47.46.2 10.2 23.7

Fig. 2.10: XRD patterns of NaX (Charnell recipe) with different amounts of dispersed CuCl.

31


a b

cFig. 2.11: SEM images of CuCl loaded samples: (a) 20 wt% CuCl/NaX, (b) 36 wt% CuCl/NaX and (c) 50 wt% CuCl/NaX.

SEM images of typical CuCl/NaX crystals are shown in Fig. 2.11a-c. For higher loadings a layer of external CuCl is observed. The sample of 20 wt% CuCl on NaX shows a relatively smooth surface. At a loading of 36 wt% a very thin layer of CuCl could be observed. At higher loading (e.g. 50 wt% in Fig. 2.11c), this layer becomes thicker. The TEM image of a small CuCl/NaX crystal is shown in Fig. 2.12. EDX analysis confirmed the presence of CuCl in the zeolite crystal. The picture shows darker spots of finely dispersed CuCl throughout the zeolite crystal. Also some external CuCl around this slightly overloaded sample is present.

Fig. 2.12: TEM image of a 50 wt% CuCl/NaX sample.

32

Chapter 2

For the NaY and 43 wt% CuCl/NaY samples, the absorption spectra of the CO stretch vibrations between 2300 – 1900 cm-1 are shown in Fig. 2.13. The adsorption of CO on NaY results in the appearance of two absorption bands at 2169 and 2120 cm-1. The dispersion of CuCl onto NaY results in the appearance of an intense absorption band at 2145-2136 cm-1 as seen in the spectrum of the 43 wt% CuCl/NaY sample. The bands at 2169 and 2120 cm-1

remain present in this spectrum as two shoulders.

23000

Wavenumber [cm-1]

Abs

orba

nce

[A.U

.]

1900200021002200

0.50

0.75

1.002145-2136

0.25

43 wt% CuCl/NaY

NaY

2169 2120

23000

Wavenumber [cm-1]

Abs

orba

nce

[A.U

.]

1900200021002200

0.50

0.75

1.002145-2136

0.25

43 wt% CuCl/NaY

NaY

2169 2120

Fig. 2.13: IR absorption spectra of CO adsorption on NaY and 43wt% CuCl/NaY at 323 K and pCO = 5 kPa.

A CO desorption experiment for the 43 wt% CuCl/NaY sample is shown in Fig 2.14. The shoulder bands at 2169 and 2120 cm-1 decrease relatively quickly and are no longer present after 30 min. The absorption band at 2145-2136 initially drops gradually and a small apparent ‘blue shift’ is observed. After 30 min the absorbance of this peak continues to decreases very slowly, but could only be removed completely at an elevated temperature (> 373 K).

23000

Wavenumber [cm-1]

Abs

orba

nce

[A.U

.]

1900200021002200

0.50

0.75

1.002145-2136

0.25

43 wt% CuCl/NaY

23000

Wavenumber [cm-1]

Abs

orba

nce

[A.U

.]

1900200021002200

0.50

0.75

1.002145-2136

0.25

43 wt% CuCl/NaY

Fig. 2.14: IR absorption spectra during CO desorption from 43 wt% CuCl/NaX by flushing with helium at 323 K. Arrow indicates direction of time advancement.

33


Fig 2.15 shows the absorption spectra of the dried (at 423 K) and the dispersed (at 623 K) 43 wt% CuCl/NaY samples after CO adsorption. At 423 K the absorption band at 2145-2136 cm-1 is very small. Dispersion of CuCl only occurs after the sample is treated at 623 K.

23000

Wavenumber [cm-1]

Abs

orba

nce

[A.U

.]

1900200021002200

0.50

0.75

1.00

0.25

623 K

423 K2169

2120

2145

-213

6

23000

Wavenumber [cm-1]

Abs

orba

nce

[A.U

.]

1900200021002200

0.50

0.75

1.00

0.25

623 K

423 K2169

2120

2145

-213

6

Fig. 2.15: IR absorption spectra for CO adsorption at 323 K and 5 kPa on mixtures of CuCl and NaY preheated at 423 or 623 K. The spectra were recorded 13 minutes after exposure to CO.

2.3.3 Adsorption of light olefins and paraffin studied with FTIR IR adsorption spectra of NaY and 43 wt% CuCl/NaY after the introduction of 3.5-5.5 kPa

of ethane or ethylene in the transmission cell are shown in Fig. 2.16. Ethylene adsorption resulted in intense absorption bands between 3124-2975 cm-1, 1926-1875 cm-1 and 1530-1380 cm-1. For ethane intense absorption bands between 3000-2890 cm-1 and 1605-1350 cm-1 were observed. As shown in the enlargement of the spectra between 1650-1250 cm-1 in Fig. 2.16d,the adsorption of ethylene on 43wt% CuCl/NaY resulted in the formation of additional bands at 1544, 1421 and 1280 cm-1, which were not observed on NaY (Fig. 2.16b).

Once the pressure in the transmission was reduced to 1.0 kPa, the intensity of the all the absorption bands decreased. For ethane a decrease in the gas pressure immediately resulted in a considerable decrease of all the absorption bands on both the NaY and the 43 wt% CuCl/NaY samples. The three additional absorption bands of ethylene on the 43 wt% CuCl/NaY sample remained relatively high. Also the absorbance at 3124-2975 cm-1 is still present at a low intensity. The absorption bands of ethylene adsorbed on the NaY sample disappeared almost completely after the reduction of the pressure. A further evacuation of the transmission cell resulted in the complete disappearance of most absorption bands. For the 43 wt% CuCl/NaY sample a slight increase in temperature could be applied to allow a faster disappearance of the absorption bands of ethylene.

34

Chapter 2

35

1500

2000

2500

3000

3500a

Wav

enum

ber[

cm-1

]

0

Absorbance[-] 0.2

0.4

C2H

6N

aY

3000

-289

0

1605

-135

0

1500

2000

2500

3000

3500a

Wav

enum

ber[

cm-1

]

0

Absorbance[-] 0.2

0.4

1500

2000

2500

3000

3500a

Wav

enum

ber[

cm-1

]

0

Absorbance[-] 0.2

0.4

C2H

6N

aY

3000

-289

0

1605

-135

0

1500

2000

2500

3000

3500b

Wav

enum

ber[

cm-1

]

0

Absorbance[-] 0.2

0.4

1300

1400

1500

1600

00.2

0.6

0.8

0.4

C2H

4N

aY

3124

-297

5

1926

-187

5

1530

-138

0

1500

2000

2500

3000

3500b

Wav

enum

ber[

cm-1

]

0

Absorbance[-] 0.2

0.4

1500

2000

2500

3000

3500b

Wav

enum

ber[

cm-1

]

0

Absorbance[-] 0.2

0.4

1300

1400

1500

1600

00.2

0.6

0.8

0.4

1300

1400

1500

1600

00.2

0.6

0.8

0.4

C2H

4N

aY

3124

-297

5

1926

-187

5

1530

-138

0

1500

2000

2500

3000

3500c

Wav

enum

ber[

cm-1

]

0

Absorbance[-] 0.2

0.4

0.6

0.8

1.0

C2H

6C

uCl/N

aY30

00-2

890

1605

-135

0

1500

2000

2500

3000

3500c

Wav

enum

ber[

cm-1

]

0

Absorbance[-] 0.2

0.4

0.6

0.8

1.0

1500

2000

2500

3000

3500c

Wav

enum

ber[

cm-1

]

0

Absorbance[-] 0.2

0.4

0.6

0.8

1.0

C2H

6C

uCl/N

aY30

00-2

890

1605

-135

0

d

bance[-]

1500

2000

2500

3000

3500

Wav

enum

ber[

cm-1

]

0Absor 0.

2

0.4

0.6

0.8

1.0

1300

1400

1500

1600

0

C2H

4C

uCl/N

aY19

26-1

875

1500

2000

2500

3000

3500

Wav

enum

ber[

cm-1

]

0Absor

0.2

0.6

0.8

0.4

1544

1280

1421

3124

-297

5

1530-1380

d

bance[-] 0.2

0.4

0.6

0.8

1.0

1500

2000

2500

3000

3500

Wav

enum

ber[

cm-1

]

0Absor

d

bance[-] 0.2

0.4

0.6

0.8

1.0

1300

1400

1500

1600

013

0014

0015

0016

00

0.2

0.6

0.8

0.4 0

C2H

4C

uCl/N

aY19

26-1

875

0.2

0.6

0.8

0.4

1544

1280

1421

3124

-297

5

1530-1380

Fig.

2.1

6: IR

abs

orpt

ion

spec

tra a

t tw

o pa

rtial

pre

ssur

es fo

r: (a

) eth

ane

on N

aY, (

b) e

thyl

ene

on N

aY, (

c) e

than

e on

43

wt%

CuC

l/NaY

, (d)

eth

ylen

e on

43

wt%

CuC

l/NaY

. (p e

than

e = 5

.5 a

nd 1

.0 k

Pa, p

ethy

lene

= 3

.5 a

nd 1

.0 k

Pa)


36

1500

2000

2500

3000

3500

Wav

enum

ber[

cm-1

]

0

Absorbance[A.U.]

a

246

C3H

8N

aY

3025-2850

1548

-133

8

1500

2000

2500

3000

3500

Wav

enum

ber[

cm-1

]

0

Absorbance[A.U.]

a

246

1500

2000

2500

3000

3500

Wav

enum

ber[

cm-1

]

0

Absorbance[A.U.]

a

246

C3H

8N

aY

3025-2850

1548

-133

8

1500

2000

2500

3000

3500

Wav

enum

ber[

cm-1

]

0

Absorbance[A.U.]

b

123

1300

1400

1500

1600

1700

00.5

1.0

1.5

C3H

6N

aY

3106-2865

1888-1826

1548

-133

817

00-1

600

1500

2000

2500

3000

3500

Wav

enum

ber[

cm-1

]

0

Absorbance[A.U.]

b

123

1500

2000

2500

3000

3500

Wav

enum

ber[

cm-1

]

0

Absorbance[A.U.]

b

123

1300

1400

1500

1600

1700

00.5

1.0

1.5

1300

1400

1500

1600

1700

00.5

1.0

1.5

C3H

6N

aY

3106-2865

1888-1826

1548

-133

817

00-1

600

1500

2000

2500

3000

3500

Wav

enum

ber[

cm-1

]

0

Absorbance[A.U.]

c

246

C3H

8C

uCl/N

aY

1548

-133

8

3025-2850

1500

2000

2500

3000

3500

Wav

enum

ber[

cm-1

]

0

Absorbance[A.U.]

c

246

1500

2000

2500

3000

3500

Wav

enum

ber[

cm-1

]

0

Absorbance[A.U.]

c

246

C3H

8C

uCl/N

aY

1548

-133

8

3025-2850

ance[A.U.]

1500

2000

2500

3000

3500

Wav

enum

ber[

cm-1

]

0Absorb

d

0.5

1.0

1.5

2.0

1300

1400

1500

1600

1700

00.5

1.0

1.5

1556

1269

1500

-135

0

3106-2865

1700-1600

ance[A.U.]

C3H

6C

uCl/N

aY

1888-1826

1500

2000

2500

3000

3500

Wav

enum

ber[

cm-1

]

0Absorb

d

0.5

1.0

1.5

2.0

1500

2000

2500

3000

3500

Wav

enum

ber[

cm-1

]

0Absorbance[A.U.] 0.

5

1.0

d1.

5

2.0

1300

1400

1500

1600

1700

00.5

1.0

1.5

1300

1400

1500

1600

1700

00.5

1.0

1.5

1556

1269

1500

-135

0

3106-2865

1700-1600

C3H

6C

uCl/N

aY

1888-1826

Fig.

2.1

7: I

R a

bsor

ptio

n sp

ectra

at

two

parti

al p

ress

ures

for

: (a

) pr

opan

e on

NaY

, (b)

pro

pyle

ne o

n N

aY, (

c) p

ropa

ne o

n 43

wt%

CuC

l/NaY

, (d

) pro

pyle

ne o

n 43

wt%

CuC

l/NaY

. (p p

ropa

ne =

5.5

and

0.1

kPa

, ppr

opyl

ene =

5.5

and

0.1

kPa

)

Chapter 2

In case the samples were exposed to 5.5 kPa of propane or propylene the absorption spectra shown in Fig 2.17 were obtained. For propane absorption bands between 3025-2850 cm-1 and 1548-1338 cm-1 were observed. For propylene absorption bands between 3106-2865 cm-1, 1888-1826 cm-1, 1700-1600 cm-1 and 1548-1338 cm-1 were seen. As shown in the enlargement of the spectra between 1750-1250 cm-1 in Fig. 2.17d, the adsorption of propylene on the 43 wt% CuCl/NaY sample resulted in the formation of additional absorption bands: one doublet at 1556 and 1269 cm-1 and multiple intense absorption bands between 1500-1350 cm-1. These additional bands were not observed on the NaY sample (Fig. 2.17b).

For propane a decrease in the gas pressure to 0.1 kPa resulted in a considerable reduction in the absorption spectra on both samples. A decrease in the pressure of propylene to 1 mbar also resulted in the almost complete disappearance of the absorption bands on the NaY sample. For the 43 wt% CuCl/NaY sample the intensity of all adsorption bands remained relatively high, especially the additional bands at 1556, 1269 cm-1 and around 1500-1350 cm-1. Only a further evacuation, assisted by a slight temperature increase, resulted in the complete desorption of propylene from the 43 wt% CuCl/NaY sample.

2.4 Discussion

2.4.1 Synthesis of zeolite NaX The XRD pattern of the NaX zeolite synthesized according to the Charnell recipe (Fig.

2.3a) corresponded well with the literature spectrum of NaX. However relatively small crystals were obtained and the porous properties (Table 2.1) were lower than the literature values. During the synthesis of these zeolite crystals nucleation occurred almost immediately after the mixing of the Si-solution with the Al-solution. This fast nucleation of many small crystals explains the smaller final average particle size. The lower values of the porous properties (SBET and Vmicro) suggest that a small amount of a nonporous impurity is present in the final product.

The zeolites synthesized following the recipe of Qiu are much larger as can been seen from the particle size distribution plotted in Fig. 2.5. The formation of additional fines during ultrasonic treatment, suggest that some of the small particles have agglomerated with each other and/or on the surface of larger crystals. In the XRD pattern of the crystals with a diameter smaller than 50 m (Fig. 2.3b) a large contribution of NaP is seen in the XRD pattern. Also a small amount of NaA can be observed in the XRD pattern. The XRD pattern of the larger crystals corresponded much better with the XRD patterns of NaX. Only a small impurity of NaA and NaP can be observed. In the SEM picture mainly the octahedral structure of Faujasite is observed. The porous properties of these larger crystals correspond well with the expected values (Table 2.1).

37


The zeolites synthesized according to the modified recipe show a narrower particle size distribution (Fig. 2.5). During ultrasonic vibrations additional fines are hardly formed. Crystals with a D50 of over 100 m were obtained. The XRD pattern of the sieve fraction below 50 m (Fig. 2.3d) agreed well with the literature for Faujasite. In this sieve fraction a small amount of NaA and NaP impurities can be found. Compared to the synthesis of Qiu (Fig. 2.3b) the intensities of the reflections of the NaP impurity are much smaller in the modified recipe. The intensity of the NaA reflections are slightly higher than those observed in the synthesis of Qiu, which suggests a small increase in the number of NaA crystals formed during the synthesis. In the SEM image of this fraction (Fig. 2.4d) the NaA impurity is visible (the cubic crystals). Fortunately these impurities are mainly located in the smaller sieve fraction. In the XRD pattern of the crystals larger than 50 m the reflections of these impurities can no longer be observed. In the SEM image (Fig. 2.4e) only the octahedral structure of Faujasite was found. Also the porous properties of this large fraction correspond well to the typical values for Faujasite (Table 2.1).

The addition of a few milligrams of Na2SiO3·9H2O in the modified recipe resulted in an increase in the formation of NaA crystals and a decrease in the formation of NaP crystals as was seen in the XRD-patterns and SEM-pictures. The octahedral crystals seem to be better developed and had a smoother surface in the modified recipe (Fig. 2.4e). Compared to Aerosil, the use of Na2SiO3·9H2O results in a much faster formation of silicate anions. In case of Aerosil particles, SiO2 first needs to hydrolyse from the particle surface, to form silicate anions. This relatively slow process results in a much slower formation of silicate anions. Therefore the Si/Al-ratio in the solution will initially be lower. The addition of a small amount of Na2SiO3·9H2O does increase this ratio in the beginning of the synthesis, which apparently reduces the nucleation rate of NaP crystals. The concentration of the silicate anions in the solution will be maintained by the gradual hydrolysis of Aerosil. A preliminary study about the effect of an increase in the surface areas of the Aerosil particles (to 200 or 380 m2 g-1) did result in smaller particles and an increase in the fraction of NaP crystals in the final product. For higher surface areas the hydrolysis will be faster resulting in a higher silicate concentration in the solution. For the growth of large NaX crystals the silicate concentration should be kept low. Also for the synthesis of other zeolites, like Mordenite, it was shown that a low silicate concentration is required to obtain large zeolites crystals (Lozano-Castello et al. (2006)). For the nucleation period the silicate concentration should apparently be slightly higher to obtain a higher purity of NaX nuclei and to suppress the formation of NaP nuclei. To this purpose the addition of a few milligrams of Na2SiO3·9H2Oproved to be beneficial.

38

Chapter 2

2.4.2 Synthesis of CuCl/Faujasite As shown in the TGA’s, the NaY and NaX zeolite lost approximately 20-21 wt% of

adsorbed water during the first temperature increase (Fig. 2.6a-b). A further temperature increase resulted in an extra mass loss of 3 wt%, ascribed to condensation of hydroxyl groups and desorption of strongly adsorbed water. The same decrease is observed for the physical mixtures.

For the physical mixtures of CuCl and Faujasite zeolites the mass of the sample remained constant at 623 K below a certain saturation loading of CuCl. For loadings above 43 wt% on NaY and 36 wt% on NaX the mass still continued to decrease at 623 K as can be seen in Fig. 2.6a-b. The excess of CuCl slowly sublimes into the flowing helium stream. This sublimation could also be observed in the colder section of the exit tube of the quartz reactor. A greenish coating of oxidized CuCl was observed on the surface of this tube section after the storage of the quartz reactor in the ambient air. With increasing heat treatment time more CuCl is able to sublime from the samples, until finally the dispersion capacity of 43 wt% for NaY (or 36 wt% for NaX) is approached (Fig. 2.7).

In the XRD patterns of the dispersed mixtures (Fig. 2.8b for NaY and Fig. 2.10 for NaX) the reflections of CuCl have almost completely disappeared below the dispersion capacity of 43 wt% for NaY and 36 wt% for NaX, while the reflections of the Faujasite zeolite still show the same intensities as for the physical mixture. This indicates that CuCl is well retained by the zeolite below these loadings and does not escape from the sample at 623 K, which is in agreement with the TGA experiments. The obvious interpretation is that CuCl has become well dispersed over the surface of the Faujasite zeolites, invisible for XRD (Xie and Tang (1990)).

When the amount of CuCl in the mixture exceeds the dispersion capacity, the reflections of crystalline CuCl do not disappear but they are only reduced in intensity after the heat treatment, indicating the presence of residual crystalline CuCl. The TGA results show that a continuous mass loss occurred during heating at 623 K when the amount of CuCl in the mixture is higher than the dispersion capacity, suggesting that heating at this temperature not only helps CuCl to disperse over the surface of the zeolite, but also helps the excess CuCl to slowly sublime into the flowing helium stream.

Although the intensities of most of the reflections of the Faujasite structure are still similar as for the physical mixture (see also Fig. 2.9b), the low angle reflections at 2 = 6.2 and 10.2 have decreased further after the dispersion of CuCl. As seen in Fig. 2.9a for the dispersed mixture, the intensity of the reflection at 2 = 6.2 is almost zero for loadings above the dispersion capacity of 43 wt%, while it is still visible for the physical mixture. These low angle reflections correspond to the pore structure of the zeolite crystal. Due to the dispersion of CuCl on the pore wall the X-ray reflection of the pores is disturbed and therefore these reflections are no longer observed.

The XRD and TGA results are in agreement with each other, since they both show a similar dispersion capacity of 43 wt% for NaY and 36 wt% for NaX. These values correspond well with the capacities reported in the literature (Xie et al. (1996)). These dispersion

39


capacities correspond to a Cu/Al ratio of 1 for NaX and 1.7 for NaY. Since CuCl will be too large to enter the sodalite cage, this corresponds to approximately 10 CuCl molecules in each super cavity of the zeolite structure. For NaX a slightly lower value is found, while NaY resulted in a slightly higher value. For zeolite NaX an interaction with the cations positioned in II or III’-sites is possible, while for NaY only an interaction with the cations in II-sites can occur. So higher CuCl loadings would have been expected on NaX. The dispersion in NaX could be affected by diffusion limitations. Since the NaX crystals were larger than the crystals in the commercial NaY sample (< 500 nm), the CuCl molecules had to diffuse over a longer distance to achieve an equal dispersion throughout the zeolite crystal. This may explain that a slightly lower dispersion capacity was found for NaX.

The SEM image of the unsaturated sample of 20 wt% CuCl/NaX (Fig. 2.11a) clearly indicates that no external CuCl is present after the heat treatment. As was shown with TGA and XRD, the CuCl did not sublime during the dispersion process and should therefore still be present in the sample. The CuCl can therefore only be located inside the pores of the zeolite. The dispersion and presence of CuCl in the zeolite crystal was confirmed with TEM (Fig. 2.12) and EDX.

Above the dispersion capacity an external presence of CuCl is observed in the SEM picture (Fig. 2.11c). Also remains of external CuCl particles have been observed. The SEM pictures confirm that only a maximum amount of CuCl can be dispersed into NaX.

During the adsorption of CO on NaY two intense absorption bands at 2169 and 2120 cm-1

are formed (Fig. 2.13). These bands can be assigned to CO interacting with the cations (Na+)of the zeolite, either linked via its carbon-atom (Na+-CO at 2169 cm-1) or via its oxygen-atom (Na+-OC at 2120 cm-1) (Knözinger and Huber (1998); Shete et al. (1998)).

After the dispersion of CuCl in NaY, the intense absorption band at 2145-2136 cm-1

appears. This band can be assigned to the adsorption of CO on the Cu+ sites (Palomino et al. (2000)). Because of the relative high intensity of this band, an overlap occurs and the absorption bands of CO adsorbed on the Na+-ions of the zeolite appear as shoulders of this intense band.

The stronger affinity of CO with the Cu+-sites can be seen in Fig 2.14. Although the absorption bands of the zeolite adsorption sites have disappeared after 30 min of flushing with helium, the Cu+-CO band is still present. CO adsorbed on the Na+-ion is only weakly bonded and will desorb relatively easily from these sites. Because of the desorption from these Na+-sites and the overlap, a small apparent blue shift of the absorption band at 2145-2136 cm-1 is observed.

The heat treatment of the physical mixture at 423 K only resulted in the drying of the material and not in the dispersion of CuCl. Only a minor rise in the absorbance is seen at 2145-2136 cm-1. At 623 K the dispersion of CuCl occurs. Due to this dispersion, much more Cu+-adsorption sites will be present on the sample treated at 623 K. This results in a much higher absorption intensity of the Cu+-CO band at 2145-2136 cm-1 (Fig. 2.15).

40

Chapter 2

2.4.3 Adsorption of light olefins and paraffins studied with FTIR Many of the absorption bands obtained in the FTIR spectra correspond to the gas phase

spectra found in the literature (Linstrom and Mallard (2005)). Since the gas is present in the transmission cell, the presence of these bands can be expected. After the decrease of the pressure in the transmission cell these vibration bands are quickly reduced. Since the paraffins are adsorbed by the weak Van der Waals forces, the absorption spectra for ethane and propane on both samples are of very weak intensity at the lower pressure (Figs. 2.16a-b and 2.17a-b).

For the olefins additional absorption bands are observed on the 43 wt% CuCl/NaY adsorbent. These bands remain present in the absorption spectra once the gas phase pressure is reduced. The bonding of ethylene with Cu+ via -complexation results in the formation of additional absorption bands in the spectra of the 43 wt% CuCl/NaX sample (Fig. 2.16d). The doublet at 1544 and 1280 cm-1 can be assigned to C=C stretching vibrations of -adsorbedethylene and the band at 1421 is assigned to CH2 scissors vibration of adsorbed ethylene (Borgard et al. (1995)). Furthermore, the band at 1926 cm-1, which can be assigned to a combination of CH2 vibrations, remains present at relatively high intensity at the lower pressure.

The propylene spectra on the 43 wt% CuCl/NaX sample showed additional adsorption bands (Fig. 2.17d). The doublet at 1556 and 1269 cm-1 can be assigned to C=C vibrations of the -adsorbed propylene and the multiple bands at 1500-1350 cm-1 are assigned to CH2

scissor vibrations (Broclawik, Kozyra, and Datka (2005); Chen et al. (1997)). Compared to ethylene the propylene molecule has an additional methyl group, and as a consequence, absorption bands around 3050-2865 cm-1, ascribed to C-H vibrations, can still be observed in the absorption spectra for the 43 wt% CuCl/NaX adsorbent at relatively high intensity at the lower pressures.

The absorption spectra clearly show that the olefins have a stronger interaction with Faujasite zeolite once CuCl is dispersed on the zeolite. Upon evacuation the absorbance of the olefins remains higher than the absorbance of the paraffin. The spectra indicate that the dispersion of CuCl results in a higher affinity of the Faujasite zeolite for the olefin compared to the paraffin. A decrease in the pressure (or an increase in the temperature) results in the complete desorption of the olefins, which suggests that a reversible adsorption process for the olefin/paraffin separation is possible.

41


2.5 Conclusions Large (~ 100 m) uniform zeolite NaX crystals with a Si/Al ratio of 1.3 were successfully

synthesized using a modified recipe. A simple and effective method was developed to disperse CuCl into the NaX crystals and a commercial NaY sample. TGA and XRD results indicate a dispersion capacity of 36 wt% CuCl onto NaX and 43 wt% onto NaY crystals, corresponding to approximately 10 CuCl molecules per supercage. TEM and SEM images revealed that the CuCl had indeed migrated into the zeolite pore structure.

The DRIFT study with CO as a probe molecule showed that CuCl was dispersed in the zeolite at 623 K. Furthermore, the adsorption of CO on Cu+ via a -complexation (donation and back-donation bonding) was relatively strong and its desorption was much slower than from the zeolite Na+-sites. CO could only be completely removed above 373 K.

The IR study of the adsorption of ethane, ethylene, propane and propylene on the samples indicates that the olefins have a larger affinity with the CuCl/Faujasite adsorbent than with the Faujasite adsorbent, while the paraffin has only a weak interaction with the adsorbents. A decrease in the gas pressure resulted in the desorption of the olefins and paraffins, which confirms that the 43 wt% CuCl/NaX zeolite is indeed a potential candidate for a selective olefin/paraffin adsorption process.

2.6 Acknowledgements The following people are acknowledged for there contribution: Arnoud Greidanus and Yan

Jiao for their preliminary studies on the synthesis of large NaX zeolite crystals and the adsorption of olefins and paraffins in the DRIFT cell, Bart van der Linden for his assistance with the low-pressure IR transmission cell, Delia van Rij for the ICP-OES and AAS analysis, Sander Brouwer for the N2-physisorption measurements, Lous Schouten for the particle size distribution, Niek van der Pers for the XRD-analysis of NaX, Patricia Kooyman for the TEM analysis and Jorge Gascon for the DRIFT-measurement with CO as a probe molecule.

2.7 List of symbols dp Particle diameter [m]pi Partial pressure of component i [Pa] SBET BET surface area [m2 g-1]Tm Melting temperature [K] Vmicro Micropore volume [cm3 g-1]

Reflection angle XRD [º]

2.8 References Baerlocher, Ch. and McCusker, L. B., Database of Zeolite Structures,

www.iza-structure.org/databases (2006). Blas, F. J., Vega, L. F. and Gubbins, K. E., Modeling New Adsorbents for Ethylene/Ethane

Separations by Adsorption via Pi-Complexation, Fluid Phase Equilibr. 150 (1998) 117-124.

42

Chapter 2

Borgard, G. D., Molvik, S., Balaraman, P., Root, T. W. and Dumesic, J. A., Microcalorimetric and Infrared Spectroscopic Studies of CO, C2H4, N2O and O2 Adsorption on Cu-Y Zeolite, Langmuir 11 (1995) 2065-2070.

Breck, D. W., Zeolite Molecular Sieves: Structure, Chemistry, and Use, Wiley-Interscience, New York (1974).

Broclawik, E., Kozyra, P. and Datka, J., IR Studies and DFT Quantum Chemical Calculations Concerning Interaction of Some Organic Molecules with Cu+ Sites in Zeolites, C. R. Chimie 8 (2005) 491-508.

Charnell, J. F., Gel Growth of Large Crystals of Sodium A and Sodium X Zeolites, J. Cryst. Growth 8 (1971) 291-294.

Chen, H. Y. and Sachtler, W. M. H., Activity and Durability of Fe/ZSM-5 Catalysts for Lean Burn NOx Reduction in the Presence of Water Vapor, Catal. Today 42 (1998) 73-83.

Chen, L., Chen, H. Y., Lin, J., Tan, K. L. and Pan, J. S., An FTIR and Static SIMS Study of the Adsorption of Propylene on Cu-ZSM-5 Catalysts, Surf. Rev. Lett. 4 (1997) 607-611.


Fitch, A. N., Jobic, H. and Renouprez, A., Localization of Benzene in Sodium-Y Zeolite by Powder Neutron Diffraction, J. Phys. Chem. 90 (1986) 1311-1318.

Grande, C. A., Firpo, N., Basaldella, E. and Rodrigues, A. E., Propane/Propylene Separation by SBA-15 and Pi-Complexated Ag-SBA-15, Adsorption 11 (2005) 775-780.




Knözinger, H. and Huber, S., IR Spectroscopy of Small and Weakly Interacting Molecular Probes for Acidic and Basic Zeolites, J. Chem. Soc. Faraday Trans. 94 (1998) 2047-2059.

Linstrom, P. J. and Mallard, W. G., NIST Chemistry WebBook, NIST Standard Reference Database Number 69, National Institute of Standards and Technology, Gaithersburg (2005).

Lozano-Castello, D., Zhu, W., Linares-Solano, A., Kapteijn, F. and Moulijn, J. A., Micropore Accessibility of Large Mordenite Crystals, Micropor. Mesopor. Mat. 92 (2006) 145-153.


Miura, H. and Gonzalez, D., A Combined Infrared and Gas Chromatographic Reaction System for in Situ Catalytic Studies, J. Phys. E. -Sci. Instrum. 15 (1982) 373-377.

Moulijn, J. A., Makkee, M. and van Diepen, A., Chemical Process Technology, John Wiley & Sons, Chichester, England (2001).

43


Olson, D. H., The Crystal Structure of Dehydrated NaX, Zeolites 15 (1995) 439-443. Padin, J., Rege, S. U., Yang, R. T. and Cheng, L. S., Molecular Sieve Sorbents for Kinetic

Separation of Propane/Propylene, Chem. Eng. Sci. 55 (2000) 4525-4535. Palomino, G. T., Bordiga, S., Zecchina, A., Marra, G. L. and Lamberti, C., XRD, XAS, and

IR Characterization of Copper-Exchanged Y Zeolite, J. Phys. Chem. B 104 (2000) 8641-8651.


Qiu, S., Yu, J., Zhu, G., Terasaki, O., Nozue, Y., Pang, W. and Xu, R., Strategies for the Synthesis of Large Zeolite Single Crystals, Micropor. Mesopor. Mat. 21 (1998) 245-251.


Robson, H. and Lillerud, K. P., Verified Synthesis of Zeolitic Materials, Elsevier, Amsterdam (2001).

Shete, B. S., Kamble, V. S., Gupta, N. M. and Kartha, V. B., Fourier Transform Infrared Study on the Encapsulation of CO in Zeolite Y under the Moderate Temperature and Pressure Conditions, J. Phys. Chem. B 102 (1998) 5581-5589.

Takahashi, A., Yang, F. H. and Yang, R. T., New Sorbents for Desulfurization by Pi-Complexation: Thiophene/Benzene Adsorption, Ind. Eng. Chem. Res. 41 (2002) 2487-2496.

Takahashi, A., Yang, R. T., Munson, C. L. and Chinn, D., Cu(I)-Y-Zeolite As a Superior Adsorbent for Diene/Olefin Separation, Langmuir 17 (2001) 8405-8414.

Van Miltenburg, A., Zhu, W., Kapteijn, F. and Moulijn, J. A., Adsorptive Separation of Light Olefin/Paraffin Mixtures, Chem. Eng. Res. Des. 84 (2006) 350-354.


Xie, Y. C. and Tang, Y. Q., Spontaneous Monolayer Dispersion of Oxides and Salts onto Surfaces of Supports : Applications to Heterogeneous Catalysis, Adv. Catal. 37 (1990) 1-43.

Xie, Y. C., Zhang, J. P., Qiu, J. G., Tong, X. Z., Fu, J. P., Yang, G., Yan, H. J. and Tang, Y. Q., Zeolites Modified by CuCl for Separating CO from Gas Mixtures Containing CO2,Adsorption 3 (1996) 27-32.

Yang, R. T., Adsorbents: Fundamentals and Applications, John Wiley & Sons, Inc., Hoboken, New Jersey (2003).

Yang, R. T. and Kikkinides, E. S., New Sorbents for Olefin Paraffin Separations by Adsorption via Pi-Complexation, AIChE J. 41 (1995) 509-517.

44

3Stability of CuCl/Faujasite adsorbents

The influence of water vapour and air on the stability of the CuCl/Faujasite adsorbent has been investigated. The exposure to ambient conditions resulted in the oxidation of the CuCl to CuCl2·3Cu(OH)2. Close to the saturation capacity of CuCl in Faujasite zeolites the effective oxidation rate appeared to be reduced. An expansion of the Cu-material was observed with SEM and TEM, attributed to the lower molar density (mol Cu per m3) of formed CuCl2·3Cu(OH)2. This resulted in the formation of fibre structures, growing perpendicular to and out of the faces of the octahedral Faujasite crystal, and in the formation of cracks in the zeolite crystals. Grooves (< 40 nm) in the grow directions of the fibres suggest an extrusion through pores. The diameter of the fibres (200-2000 nm) suggests either the merger of multiple fibres or formation of larger pores/defects on the surface of the zeolite crystal.

A DRIFTs study of CO-adsorption showed that adsorbed water resulted in a decrease of the CO adsorption capacity and/or diffusion rate. The complete desorption of water restored the original performance for CO-adsorption. Only the combined exposure to water and oxygen results in the irreversible destruction of the CuCl/Faujasite adsorbent.

TEM studies also showed a high degree of mobility of the copper species. The interaction of the CuCl/Faujasite crystal with the high energy electrons in the TEM resulted in the decomposition of CuCl into Cu-metal. Hereby Cu-metal fibres (< 50 nm) were formed, growing out of the zeolite crystal via pore openings.

Stability of CuCl/Faujasite adsorbents

3.1 Introduction Nowadays the separation of light olefin/paraffin mixtures is primarily performed via

cryogenic distillation. This process is one of the most energy intensive separation processes in the petrochemical industry (Eldridge, Siebert, and Robinson (2005); Humphrey and Keller (1997); Chapter 1 of this thesis). Therefore research is focused on finding alternative processes to achieve this separation. One of these alternatives is the development of a hybrid adsorption-distillation process. Therefore a selective adsorbent is required.

In our earlier studies (Chapter 2 of this thesis) we described the synthesis of a potential adsorbent for this separation of olefin/paraffin mixtures, namely CuCl/Faujasite. The long term storage of the samples in glass bottles in the lab indicated that an alteration of the CuCl/Faujasite samples had occurred. The original colour of the samples changed from a brownish/grey to a greenish colour. This colour change was more pronounced for the samples containing a small amount of CuCl or an overloading of CuCl. A volume expansion of the material was observed, which sometimes resulted in the doubling of the sample volume and an overflow of small open sample cups after the long time exposure to the ambient atmosphere.

For the practical application of the studied CuCl/NaX adsorbent in the industrial practice, knowledge of its stability is important. The material will have to be installed in the adsorption column and impurities can be present in the feed. It is known from the literature (Oddy and Hughes (1970)) that the exposure of CuCl to humid air could result in the oxidation of the Cu+

to Cu2+. Exposure to humid air results in the following redox-reaction:

4 CuCl + 4 H O + O CuCl ·3Cu(OH) + 2 HCl 2 2 2 2

Inside the Faujasite zeolite crystal the stability could be influenced by the pore dimensions and reduced accessibility. In this study the effect of water vapour and air on the stability of the CuCl/Faujasite adsorbent is investigated. These samples and the effect of the exposure to water and/or air are characterized with XRD, SEM, TEM, and CO adsorption studies with DRIFT analysis.

3.2 Experimental The CuCl/Faujasite zeolites were synthesized under an inert atmosphere according to our

earlier reported recipe (Chapter 2 of this thesis). To this purpose a 30 mm long ¼” stainless steel tube was filled with a physical mixture of NaX (own recipe, crystal size ~ 50-70 m) or NaY (Zeolyst, CV100, crystal size < 500 nm) and CuCl (Fluka). The material was contained within the tube by 1 mm thick stainless steel frits (pore openings of 0.5 m) at the in- and outlet of the tube. The small column was installed in a ceramic oven and an argon flow of 100 ml min-1 -1 was fed to the column continuously. The column was heated to 423 K at 1 K minwhere it remained for 1 hour. At this temperature most of the adsorbed water on the zeolite could desorb, preventing the reaction of water with CuCl at a higher temperature. Thereafter

46

Chapter 3

-1the temperature was increased to 623 K at 1 K min , where it remained for 4 hours. After that dwell time heating was stopped and the temperature slowly returned to room temperature.

After the preparation the in- and outlet of the tube were closed with valves. In a glove box, under nitrogen atmosphere, the samples were removed from the stainless steel tube and stored in bottles in the glove box for later use. ‘Exposed’ samples of the CuCl/NaX material were created by storing the samples in open (or partly closed) bottles outside the glove box for several hours (or days).

To investigate the effect of humid air on CuCl/NaX, part of a sample with a slightly overloaded amount of CuCl (40 wt%) was subjected to another treatment. The sample was installed in a quartz reactor. At room temperature a flow of air (100 ml min-1) was bubbled through water and passed through the sample bed in the quartz reactor. After the sample had completely changed to a greenish colour, it was removed from the reactor and stored in a closed bottle as the ‘oxidized’ sample.

The possibility of regenerating the ‘exposed’ samples was investigated in a Mettler Toledo TGA/SDTA851e. Therefore a physical mixture of 50 wt% CuCl and NaX was installed in a TGA sample cup. The same temperature program as used for the normal preparation was applied to obtain 50 wt% CuCl/NaX. After the preparation the sample was exposed to the ambient air for one night. The next day, the sample was again subjected to the same temperature program and a SEM picture of this ‘regenerated’ sample was taken.

XRD patterns were recorded for the ‘exposed’ sample of 40 wt% CuCl/NaX, the ‘oxidized’ 40 wt% CuCl/NaX sample, the ‘fresh’ 36 wt% CuCl/NaX sample and the NaX sample. The measurements were performed on a Bruker-AXS D5005 type diffractometer using CuK 1 radiation. SEM pictures of the NaX sample, the ‘fresh’ CuCl/NaX samples prepared in the stainless steel tube, the ‘exposed’ samples and the ‘regenerated’ sample were taken on a Philips XL20 at 15kV.

Transmission electron microscopy (TEM) was performed using a Philips CM30T electron microscope with a LaB6 filament as the source of electrons operated at 300 kV. Small CuCl/NaX crystals, prepared by subjecting a mixture of NaX and 70 wt% CuCl to the heat treatment, were mounted on a microgrid carbon polymer supported on an aluminium grid by placing a few droplets of a suspension of ground sample in n-hexane on the grid, followed by drying at ambient conditions, all in an argon glovebox. Samples were transferred to the microscope in a special vacuum-transfer sample holder under exclusion of air. EDX elemental analysis was performed using a LINK EDX system. For the TEM analysis an overloaded sample (70wt%) was used, since our initial interest was related to the remaining external CuCl. For the purpose of this study only crystals without external CuCl were investigated.

To investigate the effect of the exposure to the ambient atmosphere on the CuCl/NaX, the sample holder was removed from the TEM and was opened for 1½ hour. Thereafter the closed sample holder was reinserted in the TEM and a picture of the same zeolite crystal was recorded.

The effect of water on the adsorption of carbon monoxide (CO) on the CuCl/NaY adsorbent was investigated with a Nicolet Magna-IR 860 FTIR. Therefore three experiments were performed with a small amount of 36 wt% CuCl/NaY in the DRIFT-cell. Before the first

47


experiment the fresh sample was slowly heated in a helium flow in the DRIFT-cell to 623 K. In this way adsorbed water was removed from the sample. Thereafter it was rapidly cooled to 323 K to start the first experiment. At 323 K 5 vol% CO in helium (Hoekloos) was passed through the DRIFT cell, while the absorbance spectra of the CO adsorption were recorded for several minutes. Once the maximum absorbance was reached, CO was desorbed from the adsorbent by returning to the pure helium stream. The last remaining molecules of adsorbed CO were removed by slowly heating the sample above 373 K.

In the second experiment with the DRIFT cell, helium was bubbled through water and was passed through the gas volume around the sample holder in the DRIFT cell for 40 minutes at 323 K. Thereafter the flow was switched from this humid helium stream to 5 vol% CO in helium. Because of the slow desorption of water from the walls of the setup and from the sample, the partial pressure of water decreased very slowly, while the CO partial pressure increased quickly to its equilibrium value. The absorption spectra of the adsorption of CO (and desorption of water) were recorded during ~2½ hours. Thereafter the flow through the DRIFT cell was returned to pure helium. To remove the remaining CO and water adsorbed on the adsorbent, the temperature was slowly increased to 623 K.

Once all the adsorbed water and CO had desorbed, the temperature was rapidly decreased to 323 K. At this temperature the third experiment in the DRIFT cell was performed, in which the absorbance was recorded while the same experimental procedure as described for the first CO-adsorption experiment was followed. This way the regenerated sample could be characterized and compared with the fresh sample.

3.3 Results The XRD patterns of the ‘fresh’ (36 wt% CuCl), the ‘exposed’ (40 wt% CuCl), the

‘oxidized’ (40 wt% CuCl) and the NaX samples are shown in Fig. 3.1. For comparison the literature XRD spectra of CuCl (Nantokite) and CuCl ·3Cu(OH)2 2 (Atacamite) are included (Rruff Project (2005); Wyckoff (1963)). The XRD pattern of the fresh 36 wt% CuCl sample only shows the reflections corresponding to NaX and minor reflections at 2 = 28.5 and 47.4 corresponding to CuCl (Chapter 2 of this thesis). In the ‘oxidized’ sample the reflections corresponding to CuCl can no longer be detected and new reflections, corresponding to CuCl ·3Cu(OH)2 2, have appeared. These new reflections can also be observed in the ‘exposed’ sample at a much lower intensity.

48

Chapter 3

Rel

ativ

ein

tens

ity[-]

10 20 30 40 50

2 [-]

f) CuCl

b) Exposed 40wt% CuCl/NaX

d) NaX

c) Fresh 36wt%CuCl/NaX

e) CuCl2 · 3 Cu(OH)2

28.5

47.4

16.1 17.6

39.5

a) Oxidized 40 wt% CuCl/NaX

Rel

ativ

ein

tens

ity[-]

10 20 30 40 50

2 [-]

f) CuCl

10 20 30 40 50

2 [-]

f) CuCl

b) Exposed 40wt% CuCl/NaX

d) NaX

c) Fresh 36wt%CuCl/NaX

e) CuCl2 · 3 Cu(OH)2

28.5

47.4

16.1 17.6

39.5

a) Oxidized 40 wt% CuCl/NaX

Fig. 3.1: XRD pattern of: a) the ‘oxidized’ 40 wt% CuCl/NaX sample, b) the ‘exposed’ 40 wt% CuCl/NaX sample c) the ‘fresh’ 36 wt% CuCl/NaX sample and d) the NaX sample. The literature pattern of e) CuCl2·3Cu(OH)2 (Atacamite) and f) CuCl (Nantokite) are included for comparison (Rruff Project (2005); Wyckoff (1963)).

49


aa bb

cc Fig. 3.2: SEM pictures of the crystals of: a) NaX, b) ‘fresh’ 36 wt% CuCl/NaX, c) ‘fresh’ 50 wt% CuCl/NaX, d) ‘exposed’ 36 wt% CuCl/NaX, e) ‘exposed’ 50 wt% CuCl/NaX and f) ‘regenerated’ 50 wt% CuCl/NaX; and details of the fibres on crystals of g) ‘exposed’ 36 wt% CuCl/NaX, h) ‘exposed’ 50 wt% CuCl/NaX and i) ‘regenerated’ 50 wt% CuCl/NaX.

dd gg

ee hh

ff ii

50

Chapter 3

In Fig. 3.2 the SEM pictures of NaX and CuCl/NaX zeolite crystals are shown. The samples in Figs. 3.2b and 3.2c, containing 36 wt% or 50 wt% CuCl respectively, were stored under nitrogen in the glove box and exposure to air was minimized. The ‘exposed’ samples, with 36 wt% or 50 wt% CuCl, are pictured in Figs. 3.2d and 3.2e respectively. On both samples fibre-like structures are visible. Elemental analysis (EDX) indicated the presence of Cu, Cl and O atoms in the fibres. The ratio of these atoms varied which suggests the presence of both CuCl and oxidized CuCl. Detailed views of these fibres are shown at a higher magnification in Fig. 3.2g-h. The fibres have a ‘carpet’-like morphology but for the overloaded sample (Fig. 3.2e) also longer fibres (>50 m) have been formed. The morphology of these fibres shows straight grooves from the surface of the zeolite crystal to the end of the fibre. At the edges of the octahedral crystal, a change in the direction of these grooves on the fibres is observed in accordance with the different three-dimensional orientation of the crystal planes and zeolite pores. Due to the formation of the fibres some cracks appeared in the zeolite crystals. These cracks sometimes resulted in the breaking up of the crystal into small pieces.

Reheating the exposed sample resulted in the deformation and redistribution of these fibres as shown in Figs. 3.2f and 3.2i. Parts of the fibres were broken and spherical structures were formed on the remaining fibres and on the surface of the zeolite crystal.

A TEM picture of a small fresh CuCl/NaX crystal is shown in Fig. 3.3a. After 1½ hour exposure to the atmosphere the morphology of the CuCl in the adsorbent had changed and the picture shown in Fig. 3.3b was obtained. Clearly, the exposure to the atmosphere resulted in the formation of larger particles both in the interior and on the outside of the zeolite crystal.

aa bb

Fig. 3.3: TEM pictures of a small CuCl/NaX crystal a) before and b) after exposure to the atmosphere. (Loading CuCl < 70 wt%.)

51


aa bb

Fig. 3.4: TEM pictures of the copper fibres formed on ‘fresh’ CuCl/NaX crystals in the TEM. (Loading CuCl < 70 wt%.)

The strong electron beam of the TEM also resulted in other fibre like structures growing from the zeolite surface. Exposure of the zeolite to the electron beam for a couple of seconds resulted in the pictures shown in Fig. 3.4a-b. Although EDX analysis revealed that the interior of the zeolite crystals contained CuCl, these fibres only contained metallic copper. Instead of the formation of these fibres, sometimes the exposure to the electron beam for a few seconds resulted in the explosion of small copper particles out of the zeolite crystal.

The absorption spectra during the adsorption of 5 vol% CO in helium are shown in Fig. 3.5 for the first DRIFT-experiment. Besides the high intensity band at 2136 cm-1, two shoulder bands are seen on both sides of this band, corresponding to CO adsorbed on the Na+-sites of the zeolite (Knözinger and Huber (1998); Shete et al. (1998); Chapter 2 of this thesis). The absorption band at 2136 cm-1 increased quickly to its maximum value within the first 3 minutes. Thereafter the absorption spectrum did not change anymore. Exposure to a flow of helium resulted in the almost complete desorption of CO. A small amount could only be removed at an elevated temperature (> 373 K).

2300 2200 2100 2000 1900Wavenumber [cm-1]

0

0.5

1.0

1.5

Abs

orba

nce

[A.U

.]

2136

2300 2200 2100 2000 1900Wavenumber [cm-1]

0

0.5

1.0

1.5

Abs

orba

nce

[A.U

.]

2136

Fig. 3.5: Absorption spectra during ½ hour adsorption of 5 wt% CO in helium on the ‘fresh’ 36wt% CuCl/NaY sample at 323 K. The arrow indicates the direction of time advancement.

52

Chapter 3

After the exposure to water vapour, the spectrum of the 36 wt% CuCl/NaX sample changed. Between 3750-2500 cm

3750 3500 3000 2750 2500Wavenumber [cm-1]

0

0.25

0.75

1.00A

bsor

banc

e [A

.U.]

3250

0. 50

a

3750 3500 3000 2750 2500Wavenumber [cm-1]

0

0.25

0.75

1.00A

bsor

banc

e [A

.U.]

3250

0. 50

3750 3500 3000 2750 2500Wavenumber [cm-1]

0

0.25

0.75

1.00A

bsor

banc

e [A

.U.]

3250

0. 50

a

2300 2200 2100 2000 1900Wavenumber [cm-1]

0

0.5

1.0

1.5

Abso

rban

ce [A

.U.]

2122

2136 b

2300 2200 2100 2000 1900Wavenumber [cm-1]

0

0.5

1.0

1.5

Abso

rban

ce [A

.U.]

2122

2136

2300 2200 2100 2000 1900Wavenumber [cm-1]

0

0.5

1.0

1.5

Abso

rban

ce [A

.U.]

2122

2136 b

Fig. 3.6: Absorption spectra during 2½ hours adsorption of 5 wt% CO in helium on the humid 36wt% CuCl/NaY sample at 323 K: a) OH-region and b) Cu+-CO region. The arrow indicates the direction of time advancement.

-1 the OH-vibration bands of adsorbed water, or OH groups of the zeolite appeared on this ‘wet’ sample. In Fig. 3.6a these bands slowly decreased when the gas atmosphere in the DRIFT cell was switched to 5 vol% CO in helium (without water vapour). This decrease confirms that water is indeed desorbing from the adsorbent.

In Fig. 3.6b two bands start to appear at 2136 and 2122 cm-1. The second band at 2122 cm-1 increased from the beginning of the experiment. After 20 minutes of CO adsorption (and water desorption), only a single absorption band around 2136-2122 cm-1 was visible, which continued to increase further.

The complete desorption of water could only be achieved at higher temperatures (> 423 K). During the increase of the temperature, a small fraction of the adsorbed water reacted with the remaining CO via the water-gas-shift reaction. In the effluent of the DRIFT-cell small amounts of hydrogen and carbon dioxide were detected with a mass spectrometer attached to the exit of the DRIFT cell. After the regeneration cycle, the absorption spectra of the CO adsorption of the ‘regenerated’ sample are similar as the absorption spectra at 323 K shown earlier for the ‘fresh’ CuCl/NaX sample (Fig. 3.5).

53


In Fig. 3.7a the absorbance of the band at 2136 cm-1 is shown for the ‘fresh’ and ‘regenerated’ samples. In Fig. 3.7b the absorbance of the band at 2136-2122 cm

0 50 100 150Time [min]

0

0.5

1.0

1.5

Abs

orba

nce

[A.U

.]

b2136-2122

3354

0 50 100 150Time [min]

0

0.5

1.0

1.5

Abs

orba

nce

[A.U

.]

0 50 100 150Time [min]

0

0.5

1.0

1.5

Abs

orba

nce

[A.U

.]

b2136-2122

3354

0 5 10 15 20Time [s]

0

0.5

1.0

1.5

Abs

orba

nce

[A.U

.]

a2136

0 5 10 15 20Time [s]

0

0.5

1.0

1.5

Abs

orba

nce

[A.U

.]

a2136

-1 and the highest OH-vibration band at 3354 cm-1 during the CO-adsorption on the ‘wet’ sample are presented. Compared to the ‘fresh’ and ‘regenerated’ sample the CO-absorption band of the ‘wet’ sample showed a much slower increase over the experimental time. The absorbance at 2136-2122 cm-1 continued to increase asymptotically, while more water is desorbed from the adsorbent.

3.4 Discussion The XRD patterns presented in Fig. 3.1 indicate that the exposure of the CuCl/NaX

adsorbent to the ambient air results in the slow oxidation of CuCl towards CuCl2·3Cu(OH)2.Depending on the composition, this process occurs at a lower rate under normal storage conditions in the laboratory, which is storage in a closed bottle, flushed with nitrogen. A forced flow of humid air through a sample bed in the quartz reactor resulted in the quick oxidation of the CuCl in the adsorbent. The dispersion of CuCl in the pores of NaX does not prevent this oxidation. The faster colour change observed for the lower loadings of CuCl and the slower colour change observed close to the saturation capacity does however indicate that the rate of the oxidation reaction is reduced whenever the CuCl loading in the Faujasite zeolite is increased. The presence of CuCl in the pores and cavities of the zeolite reduces the diffusion of oxygen and water molecules and therefore the effective oxidation rate of the CuCl will be slower inside the pores of the zeolite, but will nevertheless take place.

In the case of an overloading of the NaX zeolite with CuCl, XRD characterization results showed that CuCl particles were remaining at the outside of the zeolite crystal (Chapter 2 of this thesis). Because of the absence of an additional diffusion barrier for this external CuCl, it can be much easier oxidized than the CuCl inside the zeolite. These small external CuCl

Fig. 3.7: a) Absorbance of the Cu+-CO absorption band (2136 cm-1) versus the time for the ‘fresh’ ( ) and ‘regenerated’ ( ) 36wt% CuCl/NaY sample. b) Absorbance of the CO-vibration band at 2136-2122 cm-1 ( )and the OH-vibration band at 3354 cm-1 ( ) versus time for the ‘wet’ 36wt% CuCl/NaY sample. Lines are to guide the eye.

54

Chapter 3

particles will quickly be transformed into the greenish CuCl2·3Cu(OH)2, which explains the faster colour change observed for the overloaded samples.

During the oxidation the molar density of the Cu-salt inside the pores of the zeolite decreases from 41.8 kmol Cu-ions m-3 for CuCl (Nantokite) to 33.8 kmol Cu-ions m-3 for CuCl ·3Cu(OH)2 2 (Atacamite) (Lide and Frederikse (1994)). Therefore an expansion of the dispersed material has to occur and part of this material will search for a free volume in- or outside the zeolite crystal pores. This volume expansion results in the fibres seen in the SEM pictures of the ‘exposed’ samples (Figs. 3.2d-e and 3.2g-h). Since the fibres are forced out of the pore openings of the zeolite, the morphology of these fibres will be influenced by the dimension of the pore windows and their orientation. At the edges of the crystal the three dimensional orientation of the pore openings changes and therefore a gap is seen there.

However, the diameter of the fibres (~200-2000 nm) is considerable larger than the size of the pore openings of Faujasite zeolites (0.74 nm). This could be the result of several phenomena. The fibres of multiple pores could have merged outside the zeolite crystal, a further expansion, e.g. via oxidation, may have occurred outside the zeolite, or larger pores could be formed as imperfection on the surface of the zeolite crystal. The grooves (< 40 nm apart) seen in the growth direction of the fibres suggest an extrusion out of pores. To our opinion, the fibres are formed via an extrusion through larger pores, which are formed/present as imperfections on the rough surface of the zeolite crystal. For the overloaded sample longer fibres (> 50 m) have been formed. These longer fibres are mainly located in the centre part of the eight faces of the octahedral crystals structure. This is explained by the fact that at these locations a relatively larger amount of (oxidized) Cu-salts is present in the interior of the zeolite crystal. At the edges much more pores are present for a lower amount of CuCl, resulting is many small fibres.

With TEM this oxidation of CuCl is also seen. Before the exposure of the CuCl/NaX adsorbent to the atmosphere, CuCl is finely dispersed in the zeolite and therefore no internal particles are seen at the magnification shown in Fig. 3.3a. After the exposure to the ambient atmosphere, the CuCl is oxidized and bigger particles have been formed inside and on the exterior of the small zeolite crystal. This explains the larger particles/fibres seen in Fig. 3.3b.

Reheating the normally exposed 50wt% CuCl/NaX sample to 623 K does not restore the sample to its original state. At 523 K the release of water molecules occurs according to the following reaction (Lide and Frederikse (1994)):

CuCl ·3Cu(OH) CuCl ·3CuO + 3 H O2 2 2 2

A chemical transformation with a small change in the morphology of the Cu-salts occurs on the fibres and on the surface of the zeolite crystal as can be seen on the SEM picture (Fig. 3.2i). Partial shrinkage of the fibres could be the result of the volume reduction due to the higher molar density of CuCl ·3CuO. 2

In the TEM also other fibres were formed (Anderson et al. (2005); Mayoral and Anderson (2007)). EDX showed that the decomposition of CuCl into Cu-metal occurred, probably caused by a combination of vacuum conditions and interaction with the high-energy electron

55


beam. The Cu-metal is transported out of the zeolite pores, perhaps assisted by the transport of Cl2-gas into the vacuum of the TEM setup or by the energy input of the electron beam. The rapid decomposition could even result in a sudden explosion of Cu particles out of the zeolite crystal, due to an increased pressure in the interior of the zeolite due to formation of Cl2-gasin the decomposition reaction and the initial blockage of the pore openings. The diameter of the copper fibres (~ 50 nm) is two orders of magnitude larger than the normal pore diameter of Faujasite zeolites (0.74 nm), but is the comparable with the distance of the grooves seen on the normally exposed sample in the SEM, which suggests that the fibres formed in the TEM are extruded out of similar pores as those observed in the SEM.

These results show that the copper species dispersed in the zeolite have a high degree of mobility when the sample is exposed to ambient air at room temperature or when exposed to the electron beam in the TEM.

The absorption spectra show an intense absorption band at 2136 cm-1, which is ascribed to the strong -complex of Cu+ with CO (Palomino et al. (2000)). As seen in Fig. 3.7a this adsorption is very fast for both the ‘fresh’ and the ‘regenerated’ samples. In the presence of adsorbed water (without oxygen) the initial absorbance is much lower, which suggests a slower adsorption of CO. The adsorbed water could also hinder the diffusion of CO in the pores of the zeolite or block CO adsorption on the Cu+-sites. After the complete desorption of water the CO adsorption behaviour of the ‘regenerated’ CuCl/NaY zeolite has been restored to the CO adsorption behaviour of the ‘fresh’ sample.

In the presence of adsorbed water on the adsorbent several phenomena occur during the adsorption of CO. Once CO is introduced in the sample cell the few remaining, and easily accessible Cu+-sites are quickly occupied with CO. This results in the fast initial rise of the absorption band at 2136 cm-1. Since fewer sites are available, compared to the dry sample, the intensity of this band will be relatively low. During the desorption of water more sites will become available for CO. Because of the presence of water, initially the Cu+-CO(H2O)complex could be formed (Hadjiivanov, Kantcheva, and Klissurski (1996); Kapteijn et al. (1996); Recchia et al. (2002); Sárkány (2002); Zecchina et al. (1999)). The absorption band at 2122 cm-1 could be assigned to this complex. The shift to lower wavenumbers is due to an increased charge backdonation to CO under the influence of H O.2

When water desorbed from the CuCl/NaY adsorbent, a larger number of empty Cu+-sites becomes available and therefore the absorption band at 2136 cm-1 +, corresponding to Cu -COcomplex, will continue to increase. Furthermore the dehydration of the Cu+-CO(H2O)complex occurs simultaneously and this complex is transformed into the Cu+-CO complex, explaining the slow decrease of the intensity of the absorption band at 2122 cm-1. These phenomena resulted in a ‘blue shift’ of the absorption band from 2122 cm-1 towards 2136 cm-1, as was seen in the absorption spectra of Fig. 3.6b.

The exposure of the CuCl/Faujasite samples to only water vapour did only affect its adsorption capacity, which can be restored through a simple heat treatment. The combination of both water and oxygen, like in ambient air, results in the irreversible destruction of the adsorbent. For the application of the CuCl/Faujasite adsorbent, its exposure to ambient air, in particular a gas phase containing both water vapour and oxygen, should therefore be avoided.

56

Chapter 3

3.5 Conclusions Under the prolonged exposure of CuCl/Faujasite to the atmosphere, modification of the

CuCl dispersion on Faujasite zeolites occurs. The oxygen and water vapour in the ambient air result in the oxidation of the cuprous-chloride to CuCl ·3Cu(OH)2 2. The lower molar density (mol Cu per m3) of the ‘oxidized salt’ results in an expansion of the dispersed Cu-salt out of the pores of the zeolites. Hereby fibre like structures or cracks in the crystals were formed, sometimes resulting in the breaking up of the crystals into smaller pieces. The morphology of the fibres showed grooves (< 40 nm) in the grow directions, which suggests an extrusion out of zeolite pores. The size of the fibres (200-2000 nm), however, indicates that either the merger of multiple fibres or the formation of larger pores/defects on the surface of the zeolite crystal had occurred. Probably the fibres grow out of imperfections that are formed/present on the rough surface of the zeolite crystal.

Loadings of CuCl above the saturation capacity resulted in the formation of longer fibres (> 50 m) in the centre of the faces of the octahedral crystal structure, because of the larger quantities of Cu-salt available from the interior of the zeolite crystal. After the exposure to humid air, the CuCl/Faujasite adsorbents can no longer be returned to its original state by heating. Only a dehydration of the ‘oxidized’ copper and a redistribution of the Cu-salts occurred.

The exposure of the CuCl/Faujasite adsorbent to only water vapour resulted in a decrease of the CO adsorption. This could be recovered by a simple heat treatment. For the application of the CuCl/NaX adsorbent it is therefore important to limit its exposure to both water and oxygen, like for instance in ambient air.

Besides the formation of fibres due to the oxidation of CuCl, in the TEM also the formation of copper metal fibres was observed with similar dimensions as the grooves in the SEM images. The decomposition of dispersed CuCl in the zeolite to Cu-metal was caused by a combination of vacuum conditions and interaction with the high-energy electron beam of the TEM. The Cu-metal was rapidly pushed out of large pores to the surface of the zeolite crystal, perhaps assisted by the transport of formed Cl2-gas into the vacuum of the TEM.

3.6 Acknowledgements Niek van der Pers is acknowledged for his contribution in the XRD-analysis. Patrica

Kooyman is acknowledged for her contribution in the TEM analysis.

3.7 References Anderson, P.A., Edmondson, M.J., Edwards, P.P., Gameson, I., Meadows, P.J., Johnson, S.R.,

Zhou, W.Z., Production of ultrafine single-crystal copper wires through electron beam irradiation of Cu-containing zeolite X, Z. Anorg. Allg. Chem. 631 (2005) 443-447.


57


Hadjiivanov, K. I., Kantcheva, M. M. and Klissurski, D. G., IR Study of CO Adsorption on Cu-ZSM-5 and CuO/SiO Catalysts: Sigma and Pi Components of the Cu+

2 -CO Bond, J. Chem. Soc. , Faraday Trans. 92 (1996) 4595-4600.


Kapteijn, F., Mul, G., Marban, G., Rodriguez-Mirasol, J. and Moulijn, J.A., Decomposition of Nitrous Oxide over ZSM-5 Catalysts, Stud. Surf. Sci. Catal. 101 (1996) 641-650.

Knözinger, H. and Huber, S., IR Spectroscopy of Small and Weakly Interacting Molecular Probes for Acidic and Basic Zeolites, J. Chem. Soc. Faraday Trans. 94 (1998) 2047-2059.

Lide, D. R. and Frederikse, H. P. R., CRC Handbook of Chemistry and Physics, CRC Press, Inc., Boca Raton (1994).

Mayoral, A., Anderson, P.A., Production of bimetallic nanowires through electron beam irradiation of copper- and silver-containing zeolite A, Nanotechnology 18 (2007) Art. No. 165708.

Oddy, W. A. and Hughes, M. J., The Stabilization of Active Bronze and Iron Antiquities by the Use of Sodium Sesquicarbonate, Stud. Conserv. 15 (1970) 183-189.

Palomino, G. T., Bordiga, S., Zecchina, A., Marra, G. L. and Lamberti, C., XRD, XAS, and IR Characterization of Copper/Exchanged Y Zeolite, J. Phys. Chem. B 104 (2000) 8641-8651.

Recchia, S., Dossi, C., Psaro, R., Fusi, A., Ugo, R. and Moretti, G., Dinitrogen Irreversible Adsorption on Overexchanged Cu-ZSM-5, J. Phys. Chem. B 106 (2002) 13326-13332.

Rruff Project, Database of Raman Spectroscopy, X-Ray Diffraction and Chemistry of Minerals, The University of Arizona, rruff.geo.arizona.edu/rruff (2005).

Sárkány, J., Effects of Water and Ion-Exchanged Counterion on the FTIR Spectra of ZSM-5. II. (Cu+-CO)-ZSM-5: Coordination of Cu+-CO Complex by H2O and Changes in Skeletal T-O-T Vibrations, Top. Catal. 18 (2002) 271-277.

Shete, B. S., Kamble, V. S., Gupta, N. M. and Kartha, V. B., Fourier Transform Infrared Study on the Encapsulation of CO in Zeolite Y under the Moderate Temperature and Pressure Conditions, J. Phys. Chem. B 102 (1998) 5581-5589.

Wyckoff, R. W. G., Crystal Structures 1, Interscience Publishers, New York (1963). Zecchina, A., Bordiga, S., Palomino, G. T., Scarano, D., Lamberti, C. and Salbalaggio, M.,

Mono-, Di-, and Tricarbonylic Species in Copper(I)-Exchanged Zeolite ZSM-5: Comparison with Homogeneous Copper(I) Carbonylic Structures, J. Phys. Chem. B 103 (1999) 3833-3844.

58

4Adsorption of olefins and paraffins on NaX

and CuCl modified NaX

Single component adsorption isotherms of ethane, ethylene, propane and propylene on zeolite NaX with and without a saturated (36 wt%) amount of CuCl have been investigated using the volumetric technique. The isotherms of ethylene, propane and propylene could be well described with a Dual-Site Langmuir isotherm, while for ethane the Langmuir model is adequate. The dispersion of CuCl results in a decrease of the maximum adsorption capacity of the zeolite for all components, because of the reduction of the available pore volume. For the olefins a stronger preferential adsorption via -complexation with CuCl is present, increasing the ideal adsorption selectivity considerably. While the isosteric heat of adsorption is constant for the paraffins on both adsorbents and for the olefins on NaX, on the CuCl/NaX adsorbent the olefins show a transition at around 1.7 olefin molecules complexating with CuCl per supercage, implying that ~ 17% of the Cu+ is involved in the -complexation. Application of the IAS theory, for a binary (50:50) mixture of ethylene and ethane, shows an increase in the mixture selectivity for ethylene by a factor of 10-50 after CuCl dispersion in NaX. For propylene/propane a lower increase by a factor of 2-5 is predicted. For the application in the separation of olefin/paraffin mixtures the dispersion of CuCl results in an improved adsorbent selectivity but at the expense of a reduced capacity.

Adsorption of olefins and paraffins on NaX and CuCl modified NaX

4.1 Introduction The separations of light olefin/paraffin mixtures, like ethylene/ethane and

propylene/propane, are amongst the most energy intensive separation processes. This separation, usually done by cryogenic separation, requires enormous amounts of energy for the compression of the gases and cooling to cryogenic temperatures (Eldridge, Siebert, and Robinson (2005); Humphrey and Keller (1997); Chapter 1 of this thesis). Therefore researchers are looking for alternative processes to lower the energy demand of the separation process. An interesting alternative is to perform this separation in an adsorption based process like Pressure Swing Adsorption (PSA) or Temperature Swing Adsorption (TSA) (Ruthven, Farooq, and Knaebel (1994); Thomas and Crittenden (1998)). For these processes it is of the utmost importance to find a cheap and selective adsorbent.

The difference in selectivity can be achieved using a difference in adsorption affinity. Some transition metals, like Cu+ or Ag+, can form a -complex with the double bond of the olefins (Herberhold (1974); Yang and Kikkinides (1995); Chapter 1 of this thesis). The paraffins can only adsorb by weak Van der Waals interactions with the adsorbent, and therefore a difference in adsorption affinity is obtained. This -complex should be weak enough to release the components upon a change in pressure or temperature.

In order to obtain a large number of accessible adsorption sites, the transition metals can be dispersed over supports with large surface area. Possible supports investigated by others include: ion-exchange resins (Hirai, Kurima, and Komiyama (1986); Wu et al. (1997)), -Al2O3 (Blas, Vega, and Gubbins (1998); Yang and Kikkinides (1995)), SiO2 (Padin et al. (2000); Rege, Padin, and Yang (1998)), clays (Choudary et al. (2002)), carbons (Hirai, Komiyama, and Keiichiro (1988); Mei et al. (2002)), mesoporous silica (Grande et al. (2005)) and zeolites (Pearce (1988); Takahashi, Yang, and Yang (2002)).

In this study Faujasite NaX zeolite crystals were used as support for CuCl. The zeolite was synthesized according to our recipe reported elsewhere (Van Miltenburg et al. (2006); Chapter 2 of this thesis). By thermal treatment the CuCl was dispersed in the pores of the zeolite. Because of the formation of the -complex of the olefin with CuCl, the olefin is expected to show a stronger affinity with the adsorbents after CuCl dispersion. FTIR experiments in the low-pressure transmission cell confirmed that the olefins exhibit a strong, but reversible, interaction with the CuCl/Faujasite adsorbent (Chapter 2 of this thesis). However a reduction in the pore volume of the zeolite is unavoidable, due to the presence of CuCl. This is expected to lead to a decrease in the adsorption capacity. Overall, the adsorption selectivity of the zeolite for the olefin in an olefin/paraffin mixture is expected to increase by the CuCl dispersion.

In order to be able to model the binary adsorption, single component adsorption isotherm data will be required. In this study we will determine the single component isotherms of ethane, ethylene, propane and propylene on both NaX and 36 wt% CuCl/NaX samples that were synthesized in this study, using the volumetric technique. In the literature single component adsorption isotherm data for the NaX (Faujasite) adsorbent have been reported earlier (Bezus, Kiselev, and Du (1972); Da Silva and Rodrigues (1999); Hyun and Danner

60

Chapter 4

(1982); Siperstein and Myers (2001)). The data shows a larger adsorption of the olefins over the paraffins on the NaX adsorbent. Literature, however, lacks important information about e.g. the moisture content of the sample mass used in their results, which makes it difficult to extend these data to our own synthesized NaX sample. Therefore also the adsorption isotherms of the hydrocarbons on the NaX sample will be determined, so that its performance can be compared directly with the 36 wt% CuCl/NaX sample, and can be used for the evaluation of mixture adsorption (Chapter 5 of this thesis).

Compared to the TEOM (Zhu et al. (1998)), the volumetric technique allows the measurements of isotherm data at lower pressures and since the sample is first evacuated before measurement, it is not hindered by the presence of inert gases (e.g. helium or air) inside the pores of the zeolite (Mittelmeijer-Hazeleger, et al. (2002)). The isotherm data are correlated by isotherm models, including the Langmuir and Dual-Langmuir models. These single components isotherm models are used to calculate the ideal selectivity of the adsorbents for the olefins. Based on these single component adsorption isotherms, the mixture adsorption and selectivity on NaX and CuCl/NaX are calculated for a binary (50:50) mixture of an olefin and paraffin using the Ideal Adsorbed Solution (IAS) theory (Myers and Prausnitz (1965); Chapter 1 of this thesis) and will form the basis for the interpretation of the experimental mixture adsorption results of Chapter 5 from breakthrough experiments. The effect of the CuCl dispersion in the pores of the zeolites will be discussed and compared with the results of our earlier characterization study (Chapter 2 of this thesis).

4.2 Theory In order to determine the selectivities and adsorption capacities for the modelling of the

(binary) adsorption at various temperature and pressures, the adsorption data should be described by an isotherm model. Various (semi-)theoretical models have been proposed to describe these data (Do (1998)). Ideally a combined fitting at multiple temperatures can be applied for these models, using an Arrhenius expression for the adsorption constant.

The adsorption isotherm for a single adsorption site can be described by the Langmuir expression (Langmuir (1918)),

ii

iisatii pK

pKqq1

(4.1)

where qi is the amount adsorbed, qisat is the saturation amount adsorbed, Ki is the adsorption

constant and pi is the partial pressure of component i. To get a continuous description of the temperature and pressure dependence for the adsorption amount, the adsorption constant is cast in an Arrhenius expression,

1exp 0

00 T

TTR

HTKKg

adsii (4.2)

61


where Ki(T0) is the adsorption constant at the reference temperature T0, Rg is the universal gas constant and Hads is the enthalpy of adsorption.

In case of multiple adsorption sites various modifications of the Langmuir model have been developed. In case two distinct adsorption locations are present, a Dual-Site Langmuir equation is satisfactory (Vlugt et al. (1998)),

iBi

iBisatBi

iAi

iAisatAii pK

pKq

pKpK

qq,

,,

,

,, 11

(4.3)

where qi,Asat and qi,B

sat are the saturation capacities of component i, and Ki,A and Ki,B are the adsorption constants of component i on sites A and B respectively. Both adsorption constants will change with temperature following a similar Arrhenius expression as for the Langmuir model (Eq. 4.2).

B

Besides the Langmuir and Dual-Site Langmuir isotherms, other isotherm models that that were used to correlate the experimental data are the Toth (Toth (1971)), UNILAN (Honig and Reyerson (1952)) and Virial (Barrer (1978)) models.

For the adsorption via -complexation multiple models have been applied and developed. The adsorption was either described by one of the earlier models, e.g. Langmuir (Huang, Padin, and Yang (1999)), or a combination of the UNILAN model with either the Langmuir (Yang and Kikkinides (1995)) or the Toth (Grande et al. (2005)) models. In these models the physical adsorption of the olefins and paraffins is ascribed with the Langmuir or Toth model and the adsorption constant and adsorption capacity of the olefin is assumed to be equal to those of the corresponding paraffin. The additional adsorption of the olefin via -complexation is then described by the UNILAN model. The assumption of an equal adsorption constant and equal adsorption affinity is however not (always) valid.

To describe the affinity of the adsorbent for the adsorptive at zero pressure, the Henry’s law constant can be derived. For the Langmuir and Dual-Site Langmuir adsorption isotherms the Henry’s law constant is defined by the following expression,

jji

satji

i

i

piH Kqpq

Ki

,,0, lim (4.4)

where KH,i is the Henry’s law constant, qi,jsat is the saturation amount adsorbed and Ki,j is the

adsorption constant for component i for site j. Based on the Henry’s law constants at multiple temperatures, the isosteric heat of adsorption at zero coverage (Q0

st) can be determined based on the following expression.

TK

RTK

TRQ iHg

iHg

st

1lnln ,,2

0 (4.5)

62

Chapter 4

The isosteric heat at other loadings is defined by the following expression:

qg

st

TpRQ

1ln (4.6)

So Qst can be obtained from a plot of ln p vs 1/T at constant loading. Extrapolation to zero loading should then give the same value for the isosteric heat of adsorption at zero coverage (Q0

st).These constants would give a good impression about the differences in adsorption affinity

of the two adsorbents (NaX and 36 wt% CuCl/NaX) for the olefins and paraffins.

4.3 Experimental

4.3.1 Adsorbents Faujasite NaX zeolites were synthesized according to the recipe described earlier (Van

Miltenburg et al. (2006); Chapter 2 of this thesis). In order to disperse CuCl on NaX, a physical mixtures was prepared by mixing 36 wt% CuCl (Fluka, based on the dry sample mass of NaX) with a sieve fraction (63-71 m) of the synthesized NaX crystals. This physical mixture was slowly heated (1 K min-1) in the quartz reactor to 623 K in flowing argon with a rate of 100 ml (STP) min-1 and at this temperature the samples were heated for 4 h. The samples were subsequently cooled to room temperature, removed from the reactor and stored under nitrogen in the glove box for later use.

4.3.2 Volumetric method A Micromeritics ASAP 2010 gas adsorption analyser (stainless steel version) was used to

measure the adsorption isotherms of ethane, ethylene, propane, and propylene on NaX, 36 wt% CuCl/NaX and pure CuCl in the pressure range from 0.002 to 120 kPa. The instrument was equipped with turbo-molecular vacuum pumps and three different pressure transducers (0.13, 1.33 and 133 kPa) to enhance the sensitivity in the different pressure ranges. The static–volumetric technique was used to determine the volume of the gas adsorbed at different partial pressures: upon adsorption a pressure decrease was observed in the gas phase, which is a direct measure for the amount adsorbed.

The sample tubes were loaded with NaX or 36 wt% CuCl/NaX. For the 36 wt% CuCl/NaX sample a larger amount was used, to have approximately the same sample amount of NaX. Prior to the adsorption measurements the samples were slowly outgassed in situ in vacuum for 16 h at 383 K followed by 6 h at 523 K. The evacuation at 383 K allowed (most of) the adsorbed hydrocarbons and water to escape from the sample cell at lower temperatures. This should reduce the undesired reactions of the hydrocarbons to form carbon deposits at higher temperatures, and it should prevent the possible reaction of water with CuCl. The adsorption isotherms, which will be presented in the Results section, were based on the amount of dry NaX crystals present in the sample, which was determined via a TGA analysis (Chapter 2 of this thesis). Adsorption isotherms for ethane, ethylene, propane and

63


propylene were recorded at 318, 358 and 404 K. These temperatures were considered as representative for the temperature range in which a commercial adsorptive separation process would be carried out. The samples were maintained at these temperatures using either a water bath (318 K) or an oil bath (358 and 404 K) around the sample cell. The equilibration time for each adsorption step was about 12-30 minutes for NaX, while for the CuCl/NaX sample equilibration times ranging from 12 up to 200 minutes were required.

For comparison, the adsorption of propylene on pure CuCl particles (0.2380 g) was recorded at 358 K. Before measurements the CuCl particles were first pre-treated in the quartz reactor at 623 K for 4 h in flowing argon. Thereafter the sample was installed in a sample tube and slowly outgassed.

At the end of the adsorption measurement at the highest pressure (120 kPa), two points of the desorption isotherms were recorded (at 115 and 110 kPa) to check the reversibility of the adsorption and to check whether equilibrium was reached at 120 kPa. Lower desorption pressures were also attempted. Though they did indicate a reversible adsorption/desorptionprocess, the pressure decrease seriously affected the performance of the turbomolecular pumps. Therefore no further attempts were made to record every desorption branch down to lower pressures.

At the end of each measurement the sample tube at the analysis position (containing NaX or 36 wt% CuCl/NaX) was exchanged with the other sample tube at the manual degas position (containing the other sample). Since the evacuation of the hydrocarbons from the sample cell and the desorption from the sample appeared to be a difficult task for the turbo-molecular vacuum pumps, the pump filters for the analysis and degas position were also exchanged after each analysis. The hydrocarbons seemed to dissolve partly in the vacuum oil and tend to adsorb on the molecular sieve traps before the vacuum pumps. The analysis position was therefore always attached to the vacuum pump equipped with the cleanest filters. This exchange of filters allowed us to start the analysis of another isotherm quicker, reducing the damage to the turbo-molecular vacuum pumps and enabling us to measure one isotherm every three days.

4.3.3 Adsorptives The gases used in the experiments were all supplied by HoekLoos and had the following

purities: ethane 3.0 (>99.9%), ethylene 2.8 (99.8%), propane 3.5 (>99.95%) and propylene 3.5 (99.95%).

4.4 Results

4.4.1 Isotherms The isotherm data of the adsorption of ethane, ethylene, propane and propylene on NaX

and 36 wt% CuCl/NaX are presented in Figs. 4.1a-d and 4.2a-d. The adsorption isotherm of propylene at 358 K on bulk CuCl particles is included in Fig. 4.2c for comparison. Hardly any propylene is adsorbed by bulk CuCl particles and its contribution for the 36 wt% CuCl/NaX sample can therefore be disregarded from further analysis.

64

Chapter 4

On both samples the isotherms show a larger adsorbed amount for the olefin compared to the corresponding paraffin. For ethylene/ethane the difference is larger than for propylene/propane. The adsorbed amount of all components is smaller on the 36 wt% CuCl/NaX sample (Figs. 4.1c-d and 4.2c-d) than on the NaX sample (Figs. 4.1a-b and 4.2a-b). This effect is larger for the paraffins (decrease ~ 65%) than for the olefins (decrease ~ 50%). From a comparison of the desorption branch (not shown) with the adsorption data it can be concluded that the adsorption is reversible. In all cases desorption occurred immediately when the pressure was decreased, except for ethylene, propane and propylene on 36wt% CuCl/NaX at the lower temperatures, that showed a small increase, compared with the loading at 120 kPa, indicating that equilibrium was not yet completely reached.

The experimental isotherm data at multiple temperatures were correlated by the isotherm models by non-linear least squares fitting, minimizing the sum of squared residuals of the adsorbed amounts. In order to obtain the combined fitting results, first estimates for the saturation loading(s) and adsorption constant(s) were obtained by fitting only the isotherm at 318 K. Thereafter the isotherms at 358 K and 404 K were added to the fitting procedure and all parameters were allowed to vary in this combined fitting. The reference temperature T0

was maintained at 318 K.

0

Pressure [kPa]

q[m

ol k

g-1]

102

1

5

2

3

4

10110-110-20

Pressure [kPa]

q[m

ol k

g-1]

102

1

5

2

3

4

10110-110-2

a

404 K

358 K

318 KC2H4NaX

0

Pressure [kPa]

q[m

ol k

g-1]

102

1

4

2

3

10110-10

Pressure [kPa]

q[m

ol k

g-1]

102

1

4

2

3

10110-1

b

404 K

358 K

318 K

C2H6NaX

0

Pressure [kPa]

q[m

ol k

g-1]

102

1

3

2

10110-110-20

Pressure [kPa]

q[m

ol k

g-1]

102

1

3

2

10110-110-2

c

404 K

358 K

318 K

C2H436 wt% CuCl/NaX

0

Pressure [kPa]

q[m

ol k

g-1]

102

0.5

1.5

1.0

10110-10

Pressure [kPa]

q[m

ol k

g-1]

102

0.5

1.5

1.0

10110-1

d

404 K

358 K

318 KC2H636 wt% CuCl/NaX

Fig. 4.1: Adsorption isotherms of (a) ethylene on NaX, (b) ethane on NaX, (c) ethylene on 36wt% CuCl/NaX and (d) ethane on 36 wt% CuCl/NaX at 318 K ( ), 358 K ( ) and 404 K ( ). The lines correspond to the selected model correlation. The adsorbed amounts are based on the dry amount of NaX present in the sample.

65


Table 4.1: Adsorption isotherm parameter values for ethylene, ethane, propylene and propane on NaX and 36 wt% CuCl/NaX.

qisat

[mol kg-1]qi,A

sat

[mol kg-1]Ki,A(T0)

[10-5 Pa-1]- Hads, A

[kJ mol-1]qi,B

sat

[mol kg-1]Ki,B(T0)

[10-5 Pa-1]- Hads, B

[kJ mol-1]NaX Ethylene 4.64 3.26 51.8 42.0 1.38 2.12 36.1 Ethane 4.34 4.34 2.39 30.5 Propylene 3.99 2.94 1007 51.9 1.05 13.2 51.9 Propane 3.76 2.83 62.5 43.1 0.93 1.20 43.1

36 wt% CuCl/NaX Ethylene 2.07 0.78 2355 61.2 1.29 23.0 49.3 Ethane 1.41 1.41 2.44 35.4 Propylene 1.87 1.11 1870 59.2 0.76 23.9 54.5 Propane 1.41 0.75 77.9 47.8 0.66 2.97 47.8

0

Pressure [kPa]

q[m

ol k

g-1]

102

1

5

2

3

4

10110-210-3 10-10

Pressure [kPa]

q[m

ol k

g-1]

102

1

5

2

3

4

10110-210-3 10-1

a

404 K358 K318 K

C3H6NaX

0

Pressure [kPa]

q[m

ol k

g-1]

102

1

4

2

3

10110-10

Pressure [kPa]

q[m

ol k

g-1]

102

1

4

2

3

10110-1

b

404 K358 K

318 K

C3H8NaX

0

Pressure [kPa]

q[m

ol k

g-1]

102

1

3

2

10110-110-20

Pressure [kPa]

q[m

ol k

g-1]

102

1

3

2

10110-110-2

c

404 K358 K

318 K

CuCl, 358 K

C3H636 wt% CuCl/NaX

0

Pressure [kPa]

q[m

ol k

g-1]

102

1.5

0.5

1.0

10110-10

Pressure [kPa]

q[m

ol k

g-1]

102

1.5

0.5

1.0

10110-1

d

404 K

358 K

318 K

C3H836 wt% CuCl/NaX

Fig. 4.2: Adsorption isotherms of (a) propylene on NaX, (b) propane on NaX, (c) propylene on 36wt% CuCl/NaX and (d) propane on 36 wt% CuCl/NaX at 318 K ( ), 358 K ( ) and 404 K ( ). (c) For comparison the adsorption isotherm of propylene on CuCl ( ) at 358 K is included. The lines correspond to the selected model correlation. The adsorbed amounts are based on the dry amount of NaX present in the sample.

66

Chapter 4

The Toth and UNILAN model did not result in a considerable improvement of the sum of squared residuals compared to the Langmuir or Dual-Site Langmuir model. The Virial model resulted in only a small improvement in the sum of squared residuals, but was disregarded based on the large number of empirical fitting variables. The application of the combined UNILAN-Langmuir model to describe the adsorption of the olefins via the -complex did not result in an improved fitting and/or resulted in the heterogeneity parameter (s) to be equal or close to zero. In that case the UNILAN-Langmuir model is identical to the Dual-Site Langmuir model. Furthermore the adsorption capacities and adsorption constants of the olefins and paraffins differ on NaX, which makes the assumption of equal physical adsorption for these components incorrect. Based on these results the Langmuir and Dual-Site Langmuir model were chosen to correlate the experimental data.

For ethane the Langmuir model gave a good description of the data on both adsorbents. For ethylene the Langmuir model resulted in large deviations, especially for its adsorption on 36 wt% CuCl/NaX. A better description of the isotherm data was obtained with the Dual-Site Langmuir model. The lines in Fig. 4.1a-d represent the correlation by the best describing model of each combined data set.

The isotherm data of propane and propylene were correlated to the Dual-Site Langmuir model for both adsorbents. For propane on 36 wt% CuCl/NaX (Fig. 4.2d) and for propane and propylene on NaX (Fig. 4.2a-b) the adsorption enthalpy of the two adsorption sites were kept equal. In case the adsorption enthalpy of both adsorption locations was allowed to vary independently, the fitting convergence was very slow and the outcome depended strongly on the starting values chosen. Restricting the adsorption enthalpy for these cases to a single value resulted in more realistic and less extreme values. Comparison of the adsorption enthalpies between propane and propylene or NaX and CuCl/NaX were also more realistic. Furthermore the fitting of these three sets of adsorption isotherms was hardly affected by this restriction. Just like for ethylene, two adsorption enthalpies were used for propylene on 36 wt% CuCl/NaX (Fig. 4.2c), since on this adsorbent the olefin can also adsorb via the -complex with CuCl, which is expected to result in a larger adsorption enthalpy than for the physical adsorption.

The obtained parameter estimates for all isotherms are listed in Table 4.1. If we compare these parameters no direct relation between the A-sites (or the B-sites) on the NaX and on the 36 wt% CuCl/NaX samples can be drawn, though the values for the latter adsorbent are higher than for NaX. The designation of site A or B is purely arbitrary and does not necessarily correspond to the same location on the two samples.

4.4.2 Thermodynamics The Henry’s law constants, based on the individual fit of each individual isotherm at every

temperature, were calculated by applying Eq. 4.4. The results of this calculation are listed in Table 4.2 and are plotted versus 1/RgT for ethane/ethylene and propane/propylene in Figs. 4.3a and 4.3b, respectively.

Literature values for NaX are also included in the figures (Da Silva and Rodrigues (1999); Hyun and Danner (1982)). Following Eq. 4.5 the isosteric heat of adsorption at zero coverage

67


is calculated for both the individual and the combined fitting results. In table 4.2 the result of this calculation and the literature values are listed.

The Henry’s law constants and the slopes in Fig. 4.3a-b are higher for the olefins compared to the paraffins. For ethylene the Henry’s law constants and the slope (= isosteric heat) are higher for the 36 wt% CuCl/NaX sample compared to NaX, while for ethane a decrease in the Henry’s law constant and a small increase in the slope is observed at the investigated temperatures. For propylene and propane a decrease in the Henry’s law constant and an increase in the slope is observed, when the two samples are compared.

The slopes of the Henry’s law constants (= isosteric heats) presented in Fig. 4.3a-bcorrespond well to the literature values for NaX. The isosteric heats on NaX obtained from the individual fits correspond well with the combined fitting results, though a slightly lower value, closer to the reported literature data, is obtained. For the 36 wt% CuCl/NaX sample there is a larger difference between the isosteric heat calculated from the individual fits and that from the combined fitting. The main cause for this difference is a decrease in the saturation capacities (q sat

i,j ) at the higher temperatures obtained in the individual fitting procedure, because it was allowed to vary at the individual temperatures fits, while it was assumed to be one fixed value over the entire temperature range in the combined fitting procedure.

a

0.251/RgT [mol kJ-1]

KH

[mol

kg-1

kPa-1

]

0.30 0.35 0.40

1

10

10-1

10-2

10-3

102

C2H4NaX

C2H6NaX

C2H4 36 wt% CuCl/NaX


a


KH

[mol

kg-1

kPa-1

]

0.30 0.35 0.40

1

10

10-1

10-2

10-3

102


KH

[mol

kg-1

kPa-1

]

0.30 0.35 0.40

1

10

10-1

10-2

10-3

102

C2H4NaX

C2H6NaX




KH

[mol

kg-1

kPa-1

]

0.30 0.35 0.40

1

10

10-1

10-2

10-3

102



C3H8NaX

C3H 6NaXb


KH

[mol

kg-1

kPa-1

]

0.30 0.35 0.40

1

10

10-1

10-2

10-3

102


KH

[mol

kg-1

kPa-1

]

0.30 0.35 0.40

1

10

10-1

10-2

10-3

102



C3H8NaX

C3H 6NaXb

Fig. 4.3: a) K obtained from individual fits vs. 1/RH gT for ethylene ( , ) and ethane ( , ) on NaX ( , )and 36 wt% CuCl/NaX ( , ). The solid lines represent the combined fitting results. Literature data of ethylene and ethane on NaX is represented by ( ) and ( ), respectively (Hyun and Danner (1982)). b) K obtained from individual fits vs. 1/RH gT for propylene ( , ) and propane ( , ) on NaX ( , ) and 36 wt% CuCl/NaX ( , ).The solid lines represent the combined fitting results. Literature data of propylene and propane on NaX is represented by ( ) and ( ), respectively (Da Silva and Rodrigues (1999)).

68

Chapter 4

Table 4.2: Calculated adsorption properties of the adsorptives on the adsorbents. KH, 318K

[mol kg-1 kPa-1]KH, 358K

[mol kg-1 kPa-1]KH, 404K

[mol kg-1 kPa-1]Q0

st

[kJ mol-1]Q0

st (Literature data)[kJ mol-1]

NaX Ind.* *Comb. Ethylene 1.66 0.281 0.0624 40.8 41.9 41.83d e; 27.9-40.12 ;

37.2f

Ethane 0.104 0.0270 0.00918 30.1 30.5 26,89d; 25.12-32.22e;25.9f

Propylene 26.0 3.28 0.542 48.1 51.9 42.5a b; 46.1-52.7 Propane 1.50 0.282 0.0686 38.4 43.1 35.8a c; 32.9 ; 34.4d

36 wt% CuCl/NaX Ethylene 14.2 1.70 0.192 53.4 60.9 Ethane 0.0324 0.00869 0.00238 32.4 35.4 Propylene 16.6 2.09 0.316 49.2 59.2 Propane 0.375 0.0732 0.0163 38.9 47.8 * Based on the individual fit of each isotherm at each temperature (Ind.) or based on the combined fitting of each set of isotherms at different temperatures (Comb.). a (Da Silva and Rodrigues (1999)); b (Costa et al. (1991); Ghosh, Lin, and Hines (1993); Huang et al. (1994); Järvelin and Fair (1993)); c (Loughlin, Hasanain, and Abdul-Rehman (1990)); d (Siperstein and Myers (2001)); e (Hyun and Danner (1982)); f (Bezus, Kiselev, and Du (1972)).

Fig. 4.4a-b shows the results of the isosteric heat of adsorption as a function of the loading for the two adsorbents and the four gases. At zero loading the same isosteric heats of adsorption are obtained as those presented in table 4.2. The isosteric heat remains (relatively) constant for the paraffins on both adsorbents and for the olefins on NaX. This also follows from the Langmuir or Dual-Site Langmuir adsorption modeling for these systems, cf. Table 4.1. On the 36 wt% CuCl/NaX adsorbent a transition is observed in the isosteric heat of the olefins. The transition occurs at values corresponding with the capacities for the sites with the highest heats of adsorption in the Dual-Site Langmuir modeling, ~0.78 mol/kg for ethylene and ~1.1 mol/kg for propylene (Table 4.1).

00

q [mol kg-1]

Qst

[kJ

mol

-1]

1 3 5

20

40

60

2 4

80


C2H4 NaX

C2H6 NaXC2H6 36 wt% CuCl/NaX

00

q [mol kg-1]

Qst

[kJ

mol

-1]

1 3

20

40

60

52 4

80


C2H4 NaX


00

q [mol kg-1]

Qst

[kJ

mol

-1]

1 3 5

20

40

60

2 4

80

C3H6 36 wt% CuCl/NaXC3H6 NaX


00

q [mol kg-1]

Qst

[kJ

mol

-1]

1 3 5

20

40

60

2 4

80

C3H6 36 wt% CuCl/NaXC3H6 NaX


Fig. 4.4: a) Isosteric heat of adsorption (Qst) as a function of the amount adsorbed (q) for ethylene ( , ) and ethane ( , ) on NaX ( , ) and 36 wt% CuCl/NaX ( , ).b) Isosteric heat of adsorption (Qst) as a function of the amount adsorbed (q) for propylene ( , ) and propane ( , ) on NaX ( , ) and 36 wt% CuCl/NaX ( , ).

69


4.4.3 Ideal adsorption selectivity

All adsorbents show a larger molar adsorption capacity for the olefin than for the paraffin, about 5% for NaX and 30-35% for 36 wt% CuCl/NaX. Using the estimated parameter values of the isotherm models the ideal adsorption selectivity of the adsorbent for a binary mixture can be calculated at a fixed pressure and temperature. The ideal adsorption selectivity of the adsorbent for the olefin over the paraffin is based on their single component adsorption isotherm and defined by:

1

Pressure [kPa]

Sel

ectiv

ity[-]

102

102

10

10110-1

404 K358 K318 K

aC2H4/C2H6NaX

1

Pressure [kPa]

Sel

ectiv

ity[-]

102

102

10

10110-1

404 K358 K318 K

aC2H4/C2H6NaX

1

Pressure [kPa]

Sel

ectiv

ity[-]

102

102

10

10110-1

404 K

358 K

318 K

bC2H4/C2H636 wt% CuCl/NaX

1

Pressure [kPa]

Sel

ectiv

ity[-]

102

102

10

10110-1

404 K

358 K

318 K


Fig. 4.5: Ideal adsorption selectivity for ethylene over ethane on (a) NaX and (b) 36 wt% CuCl/NaX.

1

Pressure [kPa]

Sel

ectiv

ity[-]

102

102

10

10110-1

404 K358 K318 K

aC3H6/C3H8NaX

1

Pressure [kPa]

Sel

ectiv

ity[-]

102

102

10

10110-1

404 K358 K318 K

aC3H6/C3H8NaX

1

Pressure [kPa]

Sel

ectiv

ity[-]

102

102

10

10110-1

404 K358 K

318 K


1

Pressure [kPa]

Sel

ectiv

ity[-]

102

102

10

10110-1

404 K358 K

318 K


Fig. 4.6: Ideal adsorption selectivity for propylene over propane on (a) NaX and (b) 36 wt% CuCl/NaX.

paraffin

olefinparaffinolefin q

qS / (4.6)

The limit of the ideal selectivity for high pressure is just the ratio of the saturation loading. In Fig. 4.5a-b the calculated ideal selectivity of ethylene on the NaX and the 36 wt%

CuCl/NaX samples at each analysis temperatures are plotted over the experimental pressure range. In Fig. 4.6a-b the calculated ideal selectivity of propylene for the same adsorbents are

70

Chapter 4

shown for the same temperatures and pressure range. In all figures the ideal selectivity is increased by the CuCl modification.

4.5 Discussion

4.5.1 Adsorption isotherms The adsorption isotherms of ethylene, ethane, propylene and propane on NaX and CuCl

modified NaX can be described by either the Langmuir or Dual-Site Langmuir isotherm model (Figs. 4.1a-d and 4.2a-d). For both adsorbents the adsorbed amounts of the olefins is larger than the paraffins which suggest a larger affinity of the olefins with the adsorbents. The presence of CuCl results in a reduction of the capacity, which can be ascribed to a reduction in the pore volume due to partial filling of the pores with CuCl. The reduction of the adsorbed amount for the paraffins is larger than for the olefins. The olefins benefit from the stronger -complex which they can form with the CuCl, compensating part of the reduction in adsorption capacity due to the smaller available pore space. For propylene/propane the adsorption affinity is less affected by the CuCl dispersion in NaX, though their capacities are reduced.

The models show some deviations from the experimental data. For the adsorbent where CuCl is dispersed inside the pores and cavities of zeolite NaX, these deviations occur at the higher pressures (10-100 kPa). This can be explained as follows. At these high pressures the increase in adsorbed amount is relatively small in each step of the analysis and due to the presence of CuCl in the pores of NaX, the diffusion of the adsorptives is hindered and therefore slow. As a consequence, the equilibrium between the surface and the centre of the zeolite is not yet fully established. The absence of a complete equilibration is more pronounced at the higher pressures and is also observed in the desorption branches. The slower diffusion would require a longer equilibrium time. The ASAP 2010 volumetric setup switches to a constant 10-12 minutes equilibrium time at higher pressures. Longer equilibration steps would result in larger errors due to the drift of the pressure sensors and would therefore not improve the experimental data. The presence of a larger diffusion limitation for the 36 wt% CuCl/NaX sample was also observed at lower pressures. For propane and propylene the equilibration times are even longer than for ethane and ethylene at the lower pressures, and the deviation in the desorption branch is more pronounced, especially for the 36 wt% CuCl/NaX sample. Due to the longer carbon chain, the larger diameter and the higher molecular mass, the diffusion is further reduced and becomes an even more important factor. During our earlier adsorption study in the low-pressure FTIR-transmission cell (Chapter 2 of this thesis) a short delay of a few seconds in the decrease of the additional absorption bands of adsorbed propylene could be observed on the 43 wt% CuCl/NaY sample once the pressure was reduced. For the other gases and adsorbents this delay could not be detected within the experimental timeframe (1 second) of the FTIR recordings.

Another reason for the deviation of the experimental data from the Langmuir and Dual-Site Langmuir model could be the formation of hydrocarbon deposits. The formation of deposits was mainly observed in the NaX sample with alkenes, which showed a slight brownish colour change after completing the measurement of all the adsorption isotherms. The deposits are

71


attributed to the presence of some acidic sites in the sample that probably cause polymerization (Paweewan, Barrie, and Gladden (1998); Zikanova et al. (2006)). Fortunately these deposits didn’t seriously influence the adsorption capacity, as was indicated by a re-measurement of a couple of the adsorption isotherms.

Although the general trend shows a decrease in the adsorbed amount once CuCl is dispersed in the pores of the NaX zeolite, at lower pressures this isn’t always the case. In our adsorption study in the low pressure FTIR-transmission cell (Chapter 2 of this thesis) it was shown that the presence of CuCl allowed the olefin to interact with the adsorbent via a relatively strong -complex. This complex is already formed at low pressures. In case the decrease in the adsorption capacity due to the reduction of the pore volume is less than the increase due to the formation of the -complex, the total adsorbed amount will increase. An example is for instance the increase in the adsorbed amount of ethylene at 1 kPa and 358 K which is twice as high for the 36 wt% CuCl/NaX sample compared to NaX.

The isotherm of propylene on pure CuCl particles showed only a maximum adsorption capacity of 0.075 mol kg-1 at 120 kPa and 358 K. This would correspond to 1 molecule of propylene per 135 molecules of CuCl. On the other hand for the 36 wt% CuCl/NaX sample there is 1 molecule of ethylene per 2.75 molecules of CuCl and 1 molecule of propylene per 3.0 molecules of CuCl. So, the dispersion of the CuCl into the pores of the zeolite results in a considerable increase in the number of available adsorption sites on the CuCl for -complexation. The low adsorption of propylene on pure CuCl particles indicates that the adsorption of olefins on remaining external CuCl particles can be neglected for the analysis of the results obtained for the 36 wt% CuCl/NaX adsorbent.

In table 4.1 the obtained adsorption constants for the olefins increase once CuCl is dispersed into the zeolite pores of NaX. Also the adsorption enthalpies for the olefins have increased. This clearly shows that the olefin forms a strong adsorption bond with the CuCl adsorbent via the -complex. For the paraffin the adsorption constants hardly change. A small increase in the adsorption enthalpy is observed, probably due to the increased density of the adsorbent by the coverage of the zeolite surface with CuCl, resulting in larger attractive forces (Tümsek and Inel (2003)).

The adsorption enthalpies and loadings on NaX correspond well with the data found in the literature. Adsorption enthalpies in the range of 25-32 and 27-42 kJ mol-1 were reported for ethane and ethylene, respectively. For propane and propylene isosteric heats of adsorption on NaX of 33-36 and 43-53 kJ mol-1 were reported, respectively. The combined fitting results of the Dual-Site Langmuir model shows a larger difference between the literature data and the isosteric heat of adsorption for propane on NaX, which may be the result of the larger deviation from the experimental data. The individual fits result in an isosteric heat closer to the reported range in the literature for the isosteric heat, but this fitting procedure uses a variable adsorption capacity at each temperature. A temperature dependency of the adsorption capacity is expected to be relatively small, since the pore volume would be (almost) the same at each temperature. A large change in the adsorption capacity is therefore not considered realistic.

72

Chapter 4

For the CuCl/NaX adsorbent literature data is limited. The adsorption enthalpy seems to be in the range of the data reported earlier for CuCl (Cheng and Yang (1995); Huang, Padin, and Yang (1999)). Values of the isosteric heat of adsorption for ethylene and propylene between 46-57 and 45-54 kJ mol-1 have been reported. For ethane and propane the reported values of the isosteric heat of adsorption on CuCl adsorbents are 20-23 and 22-29, respectively. The more confined space for CuCl dispersed in NaX compared to the other supports used in these literature studies is probably responsible for the somewhat higher isosteric heats of adsorption, especially for the paraffins.

4.5.2 Thermodynamics The Henry’s law constants presented in Figs. 4.3a-b were higher than those reported in the

literature, although the global trend is similar. An important aspect for an absolute comparison between our data and literature is the sample mass used in the calculations. The ‘wet’ zeolite contains approximately 23 wt% water, which results in a 30% higher uptake for the dry sample mass compared to the ‘wet’ sample mass. As was often noticed, the literature is rather vague about which sample mass was used, which makes it difficult to compare the absolute values with our data.

As mentioned earlier, the isosteric heats of adsorption for propylene on NaX (and therefore the slopes in Fig. 4.3b) are somewhat higher than the literature data shown in the figure. The other literature sources (Table 4.2) reported higher values for propylene, which correspond better to the values obtained in our results. For propane the literature sources reported consistently lower values (Table 4.2). Our value for the isosteric heat of adsorption may be affected by the larger deviation between the experimental data and the combined fitting results.

The Henry’s law constants of ethylene, presented in table 4.2 and Fig. 4.3a, are higher for the CuCl dispersed zeolite compared to the blank zeolite. For ethane the opposite trend is observed. The Henry’s law constant changes whenever the product of both the maximum adsorption capacity and the adsorption constant changes. As seen in table 4.1, the adsorption constants hardly change for ethane, though it is slightly higher for the 36 wt% CuCl/NaX adsorbent. The maximum adsorption capacity is decreased, however, due to the smaller pore volume available for adsorption. As a consequence the Henry’s law constant has decreased for ethane. The dispersion of CuCl in NaX results in an increase in the slope of the line of ethane in Fig. 4.3a and therefore a higher isosteric heat of adsorption is found for ethane on the 36 wt% CuCl/NaX adsorbent. The larger density of the adsorbent results in a small increase in the adsorption affinity of the adsorbent for ethane and therefore a stronger temperature dependence of the Henry’s law constant.

For ethylene the maximum adsorption capacity has also decreased when CuCl is dispersed into the zeolite. However, the strong interaction between ethylene and CuCl, results in a considerable increase in the adsorption constant, and the product of the two parameters results in a larger Henry’s law constant. The stronger affinity of ethylene with the CuCl adsorbent results also in a larger isosteric heat of adsorption and this explains that the slope in Fig. 4.3ais steeper compared to NaX.

73


In case of propylene, the Henry’s law constants presented in table 4.2 and Fig. 4.3b have decreased once CuCl is dispersed into the zeolite. Apparently, in this case the increase in the adsorption constants is insufficient to compensate for the effect of the reduction of the maximum saturation capacity due to the lower available pore volume. This finally results in lower Henry’s law constants for the investigated temperatures. Nevertheless, the presence of the CuCl in the pores of the zeolite still affects the isosteric heat of adsorption for propylene in the 36 wt% CuCl/NaX sample. A higher value of this isosteric heat of adsorption is found, because of the strong -complex.

Just as for ethane, the Henry’s law constants of propane decreases, the maximum adsorption capacity is lower, the adsorption constants remain relatively constant, and the isosteric heat of propane adsorption has slightly increased due to the larger attractive force.

The isosteric heat of adsorption is (relatively) constant for the paraffins on both adsorbents and for the olefins on NaX (Fig. 4.4a-b), which indicates an energetically homogeneous sorbent, for which either the Langmuir isotherm or the Dual-Site Langmuir model with equal heats of adsorption are applicable. For the olefins on 36 wt% CuCl/NaX a transition is observed, which suggest that specific interactions are involved, e.g. -complexation of the olefins with CuCl, which occurs globally around 1 mol/kg. This value corresponds on average with 1.7 olefin molecule complexating with CuCl per supercage. Since on average 10 molecules of CuCl are present per supercage of NaX, not all Cu+ is involved in the complexation.

4.5.3 Ideal adsorption selectivity The curves of the ideal adsorption selectivity of the olefins presented in Figs. 4.5a-b and

4.6a-b show a selective affinity of the adsorbents for the olefin at the lowest pressures. The Henry’s law constants of the olefins are higher than those of the paraffins, and therefore a higher selectivity at lower pressures could indeed be expected, since the Henry’s law constants describes the adsorption at zero pressure. All figures show a decrease in the ideal selectivity once the pressure is increased. At the lower pressures the selectivity is mainly determined by ratio of the Henry coefficients. Therefore higher ideal selectivities are found for the 36 wt% CuCl/NaX sample. At the higher pressure the adsorbed amount is mainly determined by ratio of the saturation capacity and will become more equal for both samples at higher pressures. An increase in the temperature causes the ideal selectivity to increase at the higher pressure range. This is due to the fact that at the higher temperature the selectivity and the decrease in the loading are more determined by the ratio of the adsorption constants, which is higher at those temperatures.

Comparing the selectivities for ethylene on the 36 wt% CuCl/NaX sample with those on the NaX sample, a large increase is observed. The dispersion of CuCl in the pores of the zeolite results in a reduction of the pore volume available for adsorption. Therefore, the capacity of both ethylene and ethane are lowered. However ethylene has a stronger affinity with the adsorbent via -complexation. The combined effect results in a larger selectivity for ethylene on the 36 wt% CuCl/NaX sample.

74

Chapter 4

For propylene only a small increase of the selectivity on 36 wt% CuCl/NaX is observed. Just as for the C2-molecules, the dispersion of CuCl in the pores of the zeolites results in a smaller volume available for the adsorption of propylene and propane. The double bond in propylene can form a -complex with CuCl, which only partly compensates the reduction in the adsorption capacity due the smaller pore volume, because of the lower adsorption affinity via the -complex compared to ethylene.

Compared to a selective separation of propane and propylene based on size exclusion or kinetic diffusion, e.g. DD3R (Zhu et al. (2000)), AlPO-14 (Cheng and Wilson (1999)), zeolite 4A (Grande, Gigola, and Rodrigues (2003)) or ITQ-3 and CHA (Olson et al. (2004)), the ideal selectivity of both adsorbents are lower, but for the NaX adsorbent the adsorbed amount is higher, the equilibration time is shorter and the diffusion is faster.

The ratio of the Henry’s law constants presented in the literature resulted in an ideal selectivity of 34 for ethylene/ethane at 293 K (Hyun and Danner (1982)) and 17 at 323 K (Bezus, Kiselev, and Du (1972)). This corresponds well with the value obtained from the isotherms of NaX obtained in our results (a ratio of the Henry’s law constant of 17 at 318 K was obtained). For propylene and propane the ratio of the Henry’s law constant of NaX results in an ideal selectivity between 17-8 for 318-404 K at zero loading, which corresponds well to the reported literature value of around 10 in the temperature range 303-473 K (Da Silva and Rodrigues (1999)).

For the 36 wt% CuCl/NaX adsorbent, the ideal selectivities (at 298 K and 100 kPa) are close to the ones reported in the literature for other CuCl adsorbents (2.1-5.3 for ethylene/ethane and 1.4-2.9 for propylene/propane at 298 K (Cheng and Yang (1995)) compared to 2.0 for ethylene/ethane and 1.5 for propylene/propane at 318 K obtained in our results). A low ideal adsorption selectivity can be expected, since at this pressure and temperature the ideal selectivity is mainly determined by the saturation capacity and not by the affinity for CuCl.

4.5.4 Mixture selectivity Through the Ideal Adsorbed Solution (IAS) theory the binary adsorption can be predicted

based on the single component isotherms (Myers and Prausnitz (1965); Chapter 1 of this thesis). The simulations for a (50:50) ethylene and ethane mixture at 318 K are plotted in Fig. 4.7a-b. The selectivities of both adsorbents remain high and are (almost) constant for lower pressures (<100 Pa). At these pressures the adsorption is mainly dominated by the affinity of the adsorbent.

For high pressures (>100 kPa) the selectivity of the adsorbent will rise with increasing pressure. At those pressures the adsorption becomes increasingly influenced by the packing efficiency of the adsorptives in the cavities and pores of the zeolite. Increasing the pressure of a gas mixture from a low to a high pressure would result in the replacement of adsorbed ethane with the smaller ethylene and the adsorption becomes affected by the entropy of adsorption.

75


In the intermediate pressure regime (100 Pa < p < 100 kPa) the adsorption of both ethane and ethylene increases and entropy effects are not yet dominating the binary adsorption. Just like the ideal selectivity (Fig. 4.7a), which only takes the affinity of the adsorbent into account, a (small) decrease is seen for the mixture selectivity of ethylene on the NaX adsorbent. For the 36 wt% CuCl/NaX adsorbent the decrease in the ideal selectivity was much larger (Fig. 4.7b), since it started at a higher value. Therefore, a more significant decrease in the mixture selectivity of ethylene is observed in the first part of the intermediate pressure regime. Once the adsorbed amount of ethylene approaches saturation, the adsorption becomes affected by the entropy and the mixture selectivity increases for the higher pressures.

Selectivity

[-]

10-3

Pressure [kPa]

q[m

ol k

g-1]

104

10

10-1

102110-2 10-1 10 103

10-2

1

102

10

1

C2H4

C2H6

a

NaX

Selectivity

[-]

10-3

Pressure [kPa]

q[m

ol k

g-1]

104

10

10-1

102110-2 10-1 10 103

10-2

1

102

10

1

C2H4

C2H6

a

NaX10-3

Pressure [kPa]

q[m

ol k

g-1]

104

10

10-1

102110-2 10-1 10 103

10-2

1

102

10

1

C2H4

C2H6

a

NaX

q[m

ol k

g-1]

10-4

Pressure [kPa]104

10

10-1

102110-2

Selectivity

[-]

10-1 10 103

10-3

1

104

103

10-2

10

102

C2H6

C2H4b

36 wt% CuCl/NaX

q[m

ol k

g-1]

10-4

Pressure [kPa]104

10

10-1

102110-2

Selectivity

[-]

10-1 10 103

10-3

1

104

103

10-2

10

102

C2H6

C2H4b

36 wt% CuCl/NaX

Fig. 4.7: Adsorbed phase loading and selectivity prediction according to IAS theory for adsorption at 318 K of a binary (50:50) mixture of ethylene and ethane as a function of pressure on (a) NaX and (b) 36 wt% CuCl/NaX.

Selectivity

[-]

10-2

Pressure [kPa]

q[m

ol k

g-1]

104

10

10-1

102110-2 10-1 10 103

1

102

10

1

C3H6

C3H8

a

NaX

Selectivity

[-]

10-2

Pressure [kPa]

q[m

ol k

g-1]

104

10

10-1

102110-2 10-1 10 103

1

102

10

1

C3H6

C3H8

a

NaX10-2

Pressure [kPa]

q[m

ol k

g-1]

104

10

10-1

102110-2 10-1 10 103

1

102

10

1

C3H6

C3H8

a

NaX

q[m

ol k

g-1]

10-3

Pressure [kPa]104

10

10-1

102110-2

Selectivity

[-]

10-1 10 103

10-2

1

103

102

10

C3H6

C3H8

b

36 wt% CuCl/NaX

q[m

ol k

g-1]

10-3

Pressure [kPa]104

10

10-1

102110-2

Selectivity

[-]

10-1 10 103

10-2

1

103

102

10

C3H6

C3H8

b

36 wt% CuCl/NaX

Fig. 4.8: Adsorbed phase loading and selectivity prediction according to IAS theory for adsorption at 318 K of a binary (50:50) mixture of propylene and propane as a function of pressure on (a) NaX and (b) 36 wt% CuCl/NaX.

The IAS theory applied to the adsorption of a binary (50:50) mixture of propane and propylene shows only a rising trend with increasing pressures, as can be seen in Fig. 4.8a-b.No drop in the selectivity can be observed at intermediate pressures. Over the presented pressure range of Fig. 4.8a-b some transitions in the slope of the selectivity versus the pressure can be observed. Just as for ethylene these transitions can be the result of the filling of the different adsorption sites on the adsorbents, e.g. on CuCl or in the zeolite pores itself.

If we compare the two adsorbents with each other an increase in the selectivity for the olefins is observed upon CuCl dispersion in NaX. For ethylene the mixture selectivity

76

Chapter 4

increases by a factor 10-50 and for propylene by a factor 2-5. The increase in the selectivityfor propylene is smaller, as was observed also for the ideal selectivity. The additional methyl group in propylene reduces its interaction with CuCl. Furthermore, the adsorption, and diffusion, of both propane and propylene is hindered more, due to the smaller pore openings and pore volumes. Therefore these values should be considered more indicative than absolute.

Compared to the ideal selectivities presented in Figs. 4.5a-b and 4.6a-b, the IAS theory calculates a higher selectivity over the whole experimental pressure range. In the case of the ideal selectivity the competition between the two components on the adsorbent with increasing pressure is neglected, resulting in a lower selectivity for the smallest and most efficiently packed component. Therefore the IAS theory results in a larger and thermodynamically more reliable selectivity prediction. In case of adsorption of different hydrocarbons in various zeolite structures the IAS theory has shown to agree excellently with molecular modelling studies (Krishna and Paschek (2000); Smit and Krishna (2003)) and was used to interpret zeolite membrane separation results (Kapteijn, Moulijn, and Krishna (2000)).

For application of the adsorbents for the separation of binary mixtures of ethylene/ethane or propylene/propane the adsorption isotherms and the calculated selectivities clearly indicate that both adsorbents would be applicable. For the zeolite NaX the selectivity is lower and the formation of some carbon deposits could result in problems over longer timescales, but the adsorption equilibrium is reached quickly, compared to the 36 wt% CuCl/NaX sample. Dispersion of CuCl in the zeolite results in a much higher selectivity, especially for ethylene. However, larger adsorption units for each separation step would be needed, because of the decreased adsorption capacity. Furthermore, the slower diffusion could be a negative factor on the commercial scale, especially if continuous membrane operation is envisaged. These problems may be reduced by the application of commercial zeolite pellets in PSA units, since these pellets usually consist of much smaller crystals than the ones used here.

4.6 Conclusions The adsorption isotherms show a preferential adsorption of the olefins on both adsorbents.

The isotherms of ethylene, propylene and propane on NaX and 36 wt% CuCl/NaX can be described by the Dual-Site Langmuir model. For ethane a single site Langmuir model is sufficient to describe the experimental data. The dispersion of CuCl in the pores and cavities of the zeolite results in lower saturation capacity of all adsorptives, but the selectivities, the adsorption constants and the isosteric heats of adsorption have increased. Analysis of the isosteric heats of adsorption as a function of the loading shows that on 36 wt% CuCl/NaX a transition occurs for the olefins at around 1.7 olefin molecules -complexating CuCl per supercage. Hence, only 17% of the CuCl is involved in the -complexation. For the olefins on NaX and the paraffins on 36 wt% CuCl/NaX and NaX the isosteric heats remains constant.

Despite the lower saturation capacity of ethylene, an increase in the Henry’s law constants is observed, since the increase in the adsorption constant of the olefin is much larger than the decrease in the saturation capacity. For propylene the decrease in the saturation capacity is only partly compensated by the increased adsorption constant. Therefore a minor decrease in the Henry’s law constants of propylene is found when CuCl is dispersed in NaX.

77


For ethane the adsorption constant hardly changes, while its saturation capacity decreases upon CuCl dispersion, resulting in smaller Henry’s law constant for the paraffin. For propane the saturation capacity decreases and the adsorption constants increases, finally resulting in smaller Henry’s law constants. Due to the coverage of the internal zeolite surface with CuCl, resulting in a smaller pore space and therefore increased attractive dispersive (Van der Waals-) force, the dispersion of CuCl results in a small increase in the isosteric heat of adsorption of the paraffins.

The measurements of the isotherms indicate that the dispersion of CuCl results in a diffusion barrier for the adsorptives. Especially for propylene and propane the measurement of the equilibrium loading at higher pressures and low temperatures are hampered by a slower diffusion.

The application of the ideal adsorbed solution (IAS) theory predicts that the binary (50:50) mixture selectivity of ethylene, once CuCl was dispersed in zeolite NaX, increases by a factor 10-50 to a value of 400 at 100 kPa and 318 K. For propylene an increase by a factor 2-5 is predicted to a value of 100 at 100 kPa and 318 K. For propylene only a small improvement in the selectivity is obtained since the additional methyl group reduces its affinity with CuCl.

For the separation of olefin/paraffin mixture the dispersion of CuCl in NaX results in an improved adsorbent selectivity. Especially for ethylene/ethane mixtures the selectivity is considerably enhanced and the dispersion of a saturated (36 wt%) amount of CuCl in NaX results in an excellent adsorbent. For the separation of propylene/propane mixtures the dispersion of CuCl results in a smaller selectivity improvement. The formation of carbon deposits seems to be reduced, but a larger diffusion hindrance is formed, which may considerably affect the industrial application of such an adsorbent for propylene/propane separation, especially for continuous membrane operation. For cyclic operations (e.g. PSA) this can be avoided if smaller crystals are used.

4.7 Acknowledgements Johan Groen and Sander Brouwer are acknowledged for their assistance with the

adsorption measurements in the volumetric setup.

4.8 List of symbols q Adsorbed amount of component i [mol kg-1]i

satq Saturation capacity for component i on site j [mol kg-1]i,j-1K Adsorption constant for component i on site j [Pa ]i,j

Rg Universal gas constant [J mol-1 K-1]T Reference temperature [K]0

H Adsorption enthalpy for site j [J mol-1]ads, j

K Henry’s law constant for component i [mol kg-1 -1 Pa ]H,i

Qst Isosteric heat of adsorption [J mol-1]st -1Q Isosteric heat of adsorption at zero coverage [J mol ]0

Solefin/paraffin Selectivity of the olefin compared to the paraffin [-] p Partial pressure of component i [Pa] i

78

Chapter 4

4.9 References Barrer, R. M., Zeolites and Clay Minerals as Sorbents and Molecular Sieves, Academic Press,

London (1978). Bezus, A. G., Kiselev, A. V. and Du, P. Q., The Influence of Size, Charge and Concentration

of Exchange Cations on the Adsorption of Ethane and Ethylene by Zeolite, J. Colloid Interface Sci. 40 (1972) 223-232.

Blas, F. J., Vega, L. F. and Gubbins, K. E., Modeling New Adsorbents for Ethylene/Ethane Separations by Adsorption via Pi-Complexation, Fluid Phase Equilibr. 150 (1998) 117-124.

Cheng, L. S. and Wilson, S. T., Vacuum Swing Adsorption Process for Separating Propylene from Propane, US Patent 6 296 688 (1999).

Cheng, L. S. and Yang, R. T., Monolayer Cuprous Chloride Dispersed on Pillared Clays for Olefin-Paraffin Separations by Pi-Complexation, Adsorption 1 (1995) 61-75.


Costa, E., Calleja, G., Jimerez, A. and Pau, J., Adsorption Equilibrium of Ethylene, Propane, Propylene, Carbon Dioxide, and Their Mixtures on 13X Zeolite, J. Chem. Eng. Data 36 (1991) 218-224.

Da Silva, F. A. and Rodrigues, A. E., Adsorption Equilibria and Kinetics for Propylene and Propane over 13X and 4A Zeolite Pellets, Ind. Eng. Chem. Res. 38 (1999) 2051-2057.

Do, D. D., Adsorption Analysis: Equilibria and Kinetics, Imperial College Press, London (1998).

Eldridge, R. B., Siebert, F. A., and Robinson, S., Hybrid Separations / Distillation Technology, Research Opportunities for Energy and Emissions Reduction, (2005).

Ghosh, T., Lin, H.-D. and Hines, A., Hybrid Adsorption-Distillation Process for Separation of Propane and Propylene, Ind. Eng. Chem. Res. 32 (1993) 2390-2399.

Grande, C. A., Firpo, N., Basaldella, E. and Rodrigues, A. E., Propane/Propylene Separation by SBA-15 and Pi-Complexated Ag-SBA-15, Adsorption 11 (2005) 775-780.

Grande, C. A., Gigola, C. and Rodrigues, A. E., Propane-Propylene Binary Adsorption on Zeolite 4A, Adsorption 9 (2003) 321-329.




Honig, J. M. and Reyerson, L. H., Adsorption of Nitrogen, Oxygen, and Argon on Rutile at Low Temperatures; Applicability of the Concept of Surface Heterogeneity, J. Phys. Chem. 56 (1952) 140-146.

79


Huang, H. Y., Padin, J. and Yang, R. T., Comparison of Pi-Complexations of Ethylene and Carbon Monoxide with Cu+ + and Ag , Ind. Eng. Chem. Res. 38 (1999) 2720-2725.

Huang, Y.-H., Johnson, J. W., Liapis, A. I. and Crosser, O. K., Experimental Determination of the Binary Equilibrium Adsorption and Desorption of Propane-Propylene Mixtures on 13X Molecular Sieves by a Differential Sorption Bed System and Investigation of Their Equilibrium Expressions, Separ. Technol. 4 (1994) 156-166.


Hyun, S. H. and Danner, R. P., Equilibrium Adsorption of Ethane, Ethylene, Isobutane, Carbon Dioxide, and Their Binary Mixtures on 13X Molecular Sieves, J. Chem. Eng. Data 27 (1982) 196-200.

Järvelin, H. and Fair, J., Adsorptive Separation of Propylene-Propane Mixtures, Ind. Eng. Chem. Res. 32 (1993) 2201-2207.

Kapteijn, F., Moulijn, J. A. and Krishna, R., The Generalized Maxwell-Stefan Model for Diffusion in Zeolites:Sorbate Molecules with Different Saturation Loadings, Chem. Eng. Sci. 55 (2000) 2923-2930.

Krishna, R. and Paschek, D., Separation of Hydrocarbon Mixtures Using Zeolite Membranes: a Modelling Approach Combining Molecular Simulations with the Maxwell-Stefan Theory, Sep. Pur. Technol. 21 (2000) 111-136.

Langmuir, I., The Adsorption of Gases on Plane Surfaces of Glass, Mica and Platinum, J. Am. Chem. Soc. 40 (1918) 1361-1403.

Loughlin, K. F., Hasanain, M. A. and Abdul-Rehman, H. B., Quaternary, Ternary, Binary and Pure Component Sorption on Zeolites. 2. Light Alkanes on Linde 5A and 13X at Moderate to High Pressures, Ind. Eng. Chem. Res. 29 (1990) 1535-1546.


Mittelmeijer-Hazeleger, M. C., Ferreira, A. F. P. and Bliek, A., Influence of Helium and Argon on the Adsorption of Alkanes in Zeolites, Langmuir 18 (2002) 9613-9616.


Olson, D. H., Camblor, M. A., Vilaescusa, L. A. and Kuehl, G. H., Light Hydrocarbon Sorption Properties of Pure Silica Si-CHA and ITQ-3 and High Silica ZSM-58, Micropor. Mesopor. Mat. 67 (2004) 27-33.

Padin, J., Rege, S. U., Yang, R. T. and Cheng, L. S., Molecular Sieve Sorbents for Kinetic Separation of Propane/Propylene, Chem. Eng. Sci. 55 (2000) 4525-4535.

Paweewan, B., Barrie, P. J. and Gladden, L. F., Coking During Ethylene Conversion on Ultrastable Zeolite Y, Appl. Catal. A-Gen. 167 (1998) 353-362.



80

Chapter 4


Siperstein, F. R. and Myers, A. L., Mixed-Gas Adsorption, AIChE J. 47 (2001) 1141-1159. Smit, B. and Krishna, R., Molecular Simulations in Zeolite Process Design, Chem. Eng. Sci.

58 (2003) 557-568. Takahashi, A., Yang, F. H. and Yang, R. T., New Sorbents for Desulfurization by Pi-

Complexation: Thiophene/Benzene Adsorption, Ind. Eng. Chem. Res. 41 (2002) 2487-2496.

Thomas, W. J. and Crittenden, B., Adsorption Technology and Design, Butterworth-Heinemann, Oxford (1998).

Toth, J., State Equations of Solid-Gas Interface Layers, Acta. Chim. Hung. 69 (1971) 311-328.

Tümsek, F. and Inel, O., Evaluation of the Thermodynamic Parameters for the Adsorption of Some N-Alkanes on A Type Zeolite Crystals by Inverse Gas Chromatography, Chem. Eng. J. 94 (2003) 57-66.


Vlugt, T. J. H., Zhu, W., Kapteijn, F., Moulijn, J. A., Smit, B. and Krishna, R., Adsorption of Linear and Branched Alkanes in the Silicalite-1, J. Am. Chem. Soc. 120 (1998) 5599-5600.


Yang, R. T. and Kikkinides, E. S., New Sorbents for Olefin Paraffin Separations by Adsorption via Pi-Complexation, AIChE J. 41 (1995) 509-517.

Zhu, W., Kapteijn, F., Moulijn, J. A., Den Exter, M. C. and Jansen, J. C., Shape Selectivity in Adsorption on the All-Silica DD3R, Langmuir 16 (2000) 3322-3329.

Zhu, W., van de Graaf, J. M., van den Broeke, L. J. P., Kapteijn, F. and Moulijn, J. A., TEOM: A Unique Technique for Measuring Adsorption Properties. Light Alkanes in Silicalite-1, Ind. Eng. Chem. Res. 37 (1998) 1934-1942.

Zikanova, A., Derewinsky, M., Sarv, P., Hudec, P., Hrabanek, P. and Kocirik, M., The Effect of Immobilization/Mobilization Processes on the Temperature Onset of a Catalyst Bed Production Studied with Ethylene Oligomerization on HZSM-5 Zeolites, Catal. Today 114 (2006) 248-256.

81

Adsorption of olefins and paraffins on NaX and CuCl modified NaX Adsorption of olefins and paraffins on NaX and CuCl modified NaX

8282

5Binary adsorption of ethylene/ethane and propylene/propane mixtures on NaX and

CuCl/NaX

The binary separation of mixtures of ethylene/ethane and propylene/propane has been investigated on NaX with and without a saturation loading (36 wt%) of dispersed CuCl. Breakthrough and desorption profiles were recorded at 318, 358 and 408 K for binary (50:50) mixtures (in helium) with partial pressures of the two components between 0.8-54 kPa. The breakthrough profiles indicate that the dispersion of CuCl in the pores of NaX resulted in a slower diffusion of the components in and out of the zeolite crystal. The breakthrough profiles show that a displacement of adsorbed paraffins by the olefins occurs along the column length. This roll-up phenomenon was more pronounced for the NaX zeolite without CuCl, where in the initial phase of the adsorption process a considerable uptake of the paraffins occurred.

The determined mixture loadings on NaX corresponded well with the prediction of the IAS-theory. The adsorption of propylene and propane on NaX showed a transition from an enthalpy controlled adsorption at lower loadings to an entropy affected adsorption at higher loadings. A similar transition was observed for ethylene and ethane on CuCl/NaX. The mixture selectivity for the olefins on NaX remained in the range of 3-10 over the entire pressure range. The dispersion of CuCl on NaX increased the mixture selectivity of the adsorbent to 15-200 for ethylene and 15-30 for propylene. The observed mixture adsorption for the CuCl/NaX adsorbent did not properly compare with the IAS-prediction, especially for propane and propylene, though qualitatively the trends corresponded well.

Binary adsorption of ethylene/ethane and propylene/propane mixtures on NaX and CuCl/NaX

5.1 Introduction The separation of light olefin/paraffin mixtures is one of the most energy intensive process

in the chemical industry (Eldridge, Siebert, and Robinson (2005); Humphrey and Keller (1997); Chapter 1 of this thesis). As an alternative for the cryogenic distillation of these gases, an adsorbent can be used to perform (part of) this separation process via adsorption. Therefore several adsorbents have been investigated including zeolites. Earlier we optimized the synthesis of two potential adsorbents for the selective adsorption of the olefins, namely NaX and CuCl/NaX (Van Miltenburg et al. (2006); Chapter 2 of this thesis).

The adsorption of ethylene, ethane, propylene and propane on the NaX zeolite and on the CuCl dispersed NaX zeolite were tested earlier by measuring single component adsorption isotherms (Chapter 4 of this thesis). The dispersion of CuCl in the NaX zeolite increased the adsorption affinity of the zeolite for the olefins. The olefins can form a -complex with the Cu+-ion (Herberhold (1974); Chapter 1 of this thesis), while the paraffins can only adsorb by a weak physical adsorption.

For the practical separation of the gases and for equipment design, binary adsorption data are required. Compared to the measurement of single component isotherm data, the measurement of binary adsorption data is more cumbersome or requires an accurate description of the density of the gas-phase (Keller and Staudt (2005)). In case the static volumetric or gravimetric method is used, the composition of the gas phase has to be determined after the equilibrium has been reached in the sample cell for instance by gas chromatography (Al-Baghli and Loughlin (2006); Ghosh, Lin, and Hines (1993); Lee et al. (2004); Linders et al. (2001)). This sampling of the head-space can disturb the next steps in the volumetric measurement. Especially since a relatively small gas volume will be preferred in an optimized volumetric setup, since otherwise large amounts of adsorbent will be required to obtain accurate measurements of the pressure decrease and thus the adsorbed amount. Because of this small gas volume, the sampling volume for the analysis is relatively large and can seriously affect the accuracy of the measurements of the subsequent steps in the volumetric analysis. Furthermore the final gas phase composition cannot be controlled a priori, and is a result of the uptake process.

Instead of the analysis of the gas phase composition via gas chromatography, the composition can also be determined by the measurement of the density of the gas phase (Keller, Iossifova, and Zimmermann (2005)). This technique would require an accurate description of the density of the gas phase as a function of composition, temperature and pressure and is questionable for adsorption of isomers. Also this approach does not allow the control of the final gas phase composition at equilibrium.

An alternative is the use of dynamic techniques. The binary gas mixture will flow through (or along) the sample bed and the temperature and pressure in the bed are maintained at a fixed setting. Compared to the static methods, the flow of the gas through the sample bed allows a good contact with the adsorbent and reduces external mass transfer limitations. To determine the amount and the composition on the adsorbent, the feed and outlet flows have to be monitored continuously, otherwise only the total uptake will be obtained (Peng et al.

84

Chapter 5

(2005)). In the Zero Length Column (ZLC) technique (Brandani and Ruthven (2003)) the desorption of a binary mixture from a small amount of adsorbent is monitored versus time. Based on these response curves the adsorbed amount of both components can then be obtained and the selectivity can be calculated. This ZLC technique is mainly applied in the low concentration regime and for the measurement of diffusion data.

Monitoring the response of a packed column of the adsorbent, which is subjected to a concentration pulse, allows the calculation of binary adsorption data using the retention time of this response pulse from the packed chromatographic column (Denayer and Baron (1997); Harlick and Tezel (2000)). The presence of the second component will affect the interaction of the gases with the adsorbent. The reduced (or enhanced) interaction results in shorter (or longer) retention time and allows the calculation of the binary adsorption isotherm from transient model equations.

Alternatively the column can be subjected to a step change in the composition or pressure (Bárcia, Silva, and Rodrigues (2006); Sakuth, Meyer, and Gmehling (1998)). Increasing the partial pressure of the components will yield breakthrough profiles. Returning to an inert carrier gas, without any of the adsorptives, will yield the desorption profiles of loaded breakthrough columns. Besides the use of a carrier/sweep gas, switching from the adsorbing to the desorbing state of the breakthrough column can also be achieved by an increase in the temperature (TSA) or by a decrease of the total pressure (PSA). To be able to monitor the adsorption and desorption profiles continuously within the experimental timeframe of these profiles a fast analysis technique is required.

In the industrial practice the operation of an adsorption based process (e.g. PSA or TSA) very closely resembles the experimental procedure used in breakthrough experiments. Because of the close approximation to the real industrial process, various models have been developed to describe the kinetics in the breakthrough columns for single or multi-components systems (Grande and Rodrigues (2005); Krishna and Baur (2003)). These models include for instance the effects of diffusion in the zeolite crystal and the effects of the adsorption and desorption of all components on the Maxwell-Stefan diffusion modelling.

Table 5.1: Comparison of various techniques to measure binary adsorption. Breakthrough

column Chromato-

graphic column

Zerolength

column

Volumetric/ gravimetric

+ chromatography

Volumetric/ gravimetric

+ densimetry Gas phase Dynamic Dynamic Dynamic Static StaticGood contact gas-sorbent

Yes Yes Yes No No

Control of gas phase composition

Yes Yes Yes No No

Accurate model of setup/gas required

No Yes No No Yes

Complexity of setup* 0 + +/0 - 0/- Concentration range Low-high Low-high Low Low-high Low-high Industrial resemblance

Yes Yes/No No No No

* ‘+’ means more complex, ‘-’ means less complex, ‘0’ is the reference.

85


The comparison of the various techniques to measure binary adsorption data is summarized in Table 5.1. Because of the close approximation to the industrial practice, their ease of operation and exact control of the binary gas phase composition at equilibrium, breakthrough and desorption experiments were chosen as the experimental method to obtain the binary adsorption data. The breakthrough and desorption profiles allow an immediate screening of potential adsorbents and operating conditions for a good separation of light olefin/paraffin mixtures. In this study the performance of the NaX zeolite, with or without dispersed CuCl, was investigated for the separation of ethylene/ethane or propylene/propane mixtures. Therefore a breakthrough setup was built and multiple breakthrough and desorption profiles were recorded at three temperatures and various partial pressures for a 50:50 olefin/paraffin mixture. Based on the breakthrough and desorption profiles, the binary adsorption data were calculated using simple mass balance equations. The outcome is compared with the Ideal Adsorbed Solution (IAS) theory (Myers and Prausnitz (1965); Chapter 1 of this thesis) using the single component adsorption isotherms recorded earlier (Chapter 4 of this thesis). This theory takes into account that a competition between the two components will occur during binary adsorption. In the ‘Ideal’ solution it is assumed that each component in the adsorbed phase has an activity coefficient equal to one. This theory is also thermodynamically consistent in case the saturation capacities of both components in the mixture are not identical, which is not the case for the frequently used multicomponent Langmuir model. Although molecular modelling techniques, like configurational-bias Monte Carlo (Smit and Krishna (2003)), would make it possible to predict the binary adsorption data and these modelling techniques become increasingly powerful and successful in describing adsorption equilibria, they do not yet have a predictive reliability, especially if different types of interactions play a role (e.g. Van der Waals and -complexation for CuCl/Faujasite). Calibration of these simulations against experimental data still remains essential.

5.2 Mass balances To determine a single point of a binary adsorption isotherm from the breakthrough or

desorption profiles an exact description of the kinetics and diffusion in the breakthrough column is not required. As long as equilibrium is reached at the end of a breakthrough experiment, the loadings of both components can be determined based on the individual and total component balances over the breakthrough column during the time of the analysis. The kinetic and diffusion phenomena will however affect the shape of the breakthrough and desorption profiles. In this study no attempt is made to model these phenomena and to describe the shape of the breakthrough profiles.

In a breakthrough experiment, a gas mixture of known composition and flow is fed to the adsorption column. In the column (part of) the gas is adsorbed on the adsorbent. The remaining gas escapes from the exit of the column, where its composition is analyzed. Due to the adsorption of the components, the velocity of the gas leaving the column will be smaller compared to the inlet velocity assuming constant pressure and temperature. To calculate the adsorbed amount of each component, the outlet flowrate and composition will be required. A continuous, direct measurement of the outlet flow will be a cumbersome procedure. The

86

Chapter 5

values obtained by digital gas flow sensors are a function of the composition of the gas. Unfortunately the composition of the gas will change during a breakthrough or desorption experiment. In order to obtain reliable velocity data, an extensive calibration of the flow sensor would be required.

Fortunately the exit flow of the column can be calculated based on the molar balances across the breakthrough column. The overall mass balance at any given time during the breakthrough experiment is given by (Malek et al. (1995); Malek and Farooq (1997)):

N

i

i

g

totoutvinv t

qVTR

P1

,, 1 (5.1)

The total pressure of the breakthrough column remains constant and is set at Ptot = 108 kPa for most of the experiments in this study. The pressure drop across the column is neglected in this equation. Furthermore isothermal operation is assumed. In case of the adsorption of a binary mixture, N equals 2. For the adsorbing components the following transient balance is valid:

tq

Vtx

TRVP

TRPxx ii

g

tot

g

totoutioutviniinv 1,,,, (5.2)

Combining these two equations and rearranging, gives an expression for the outlet flow ( v, out).

N

iouti

N

i

iN

iiniinv

outv

x

tx

Vx

1,

11,,

,

1

1 (5.3)

In case the change in the outlet flow is mainly caused by the adsorption of the gases on the

adsorbent, the term N

i

i

tx

V1

, which represents the uptake and release of component i in the

void space between the adsorbent particles in the column ( V), can be neglected from the equation.

87


The outlet flow will be related to the inlet flow and the in- and outlet composition by the following equation:

N

iouti

inheliumvN

iouti

N

iiniinv

outv

xx

x

1,

,,

1,

1,,

,

11

1 (5.4)

where helium is used as an inert, non-adsorbing carrier gas. The total amount of component i adsorbed on the column can then be calculated by the

following integral starting from the start time of the experiment (tstart):

startstart tN

iouti

outiinheliumviniinv

moltoutioutviniinv

moli dt

x

xx

Vdtxx

Vq

1,

,,,,,,,,,

1

11 (5.5)

where Vmol is the molar volume of the gas at the same conditions as the volumetric flowrates. For the desorption, the inlet mol fractions of the adsorbed components (xi,in) are zero and in

this study only helium enters the column. This finally results in a similar expression as Eq. 5.4 for the adsorption experiments:

N

iouti

inheliumvN

iouti

invoutv

xx1

,

,,

1,

,,

11 (5.6)

The amount of component i desorbed from the column can then be calculated by the following integral starting from the start time of the experiment (tstart), which should ideally result in the same loading as calculated form the breakthrough profile and Eq. 5.5:

startstart tN

iouti

outiinheliumv

moltoutioutv

moli dt

x

xV

dtxV

q

1,

,,,,,

1

11 (5.7)

In case a breakthrough experiment is performed without the presence of the carrier gas (helium in this case), the outlet flow cannot be calculated from Eq. 5.4. Although initially the mol fractions of both components will be zero, after the breakthrough of the first component, the sum of the mol fractions of the two components will always be (close to) one. As a consequence the denominator becomes zero. A small experimental error would yield

88

Chapter 5

unrealistic high and/or negative outlet flows, resulting in an infinite and/or negative adsorbed amount of the components on the adsorbent.

Furthermore, the absence of a carrier gas results in a faulty analysis of the composition in the first section of the breakthrough profile. Since both components are adsorbed on the column, the outlet flow will initially be zero and no gas will leave the column towards the analysis section. In our breakthrough setup, the column was always operated (slightly) above atmospheric pressure (108 kPa), while the analysis was performed at atmospheric pressure. Because of the pressure difference some of the gas can still leave the breakthrough column via the sample line during the initial part of the breakthrough experiment. Since all the gas that is fed to the column will initially be adsorbed, the pressure of the breakthrough column will slowly decrease to atmospheric pressure. Once the breakthrough column is at atmospheric pressure, all the gas will remain inside the breakthrough column and the flow out of the column and to the analysis will be zero. The analyzed samples will then give a false representation of the actual composition at the exit of the breakthrough column and the adsorption process, since the gas in the sample loops will not be replaced by a fresh sample. After the breakthrough of the first component, the pressure will slowly start to increase towards its set-point (108 kPa). Because of this increase, the pressure difference between the breakthrough column and the analysis section will increase and the gas in the sample loops of the gas chromatograph will be refreshed.

These errors in the analysis limit the use of the breakthrough profiles of pure gas (mixtures) for a qualitative interpretation. Such profiles cannot be used to calculate the adsorbed amounts. At the end of the breakthrough experiment equilibrium will be reached and the pressure has returned to the original set-point. The desorption profiles will not be hampered by the lack of flow, since helium will be fed to the column as an inert non-adsorbing sweep gas. Therefore binary adsorption data of these pure gas mixtures will only be calculated based on the desorption profiles with the aid of Eq. 5.7.

These aspects clearly indicate the necessity to use a non-adsorbing component in a mixture as a kind of internal standard for quantitative analysis of breakthrough experiments.

5.3 Experimental

5.3.1 Breakthrough setup The adsorption of binary mixtures of ethylene/ethane or propylene/propane in helium was

investigated in a newly constructed breakthrough setup. A flowsheet of the setup is shown in Fig. 5.1.

The setup consists of a gas mixing and flow control section, a possibility for liquid injection and evaporation, the breakthrough column within a ceramic oven, several backpressure controllers and the analysis section. Part of the setup is built inside a convection oven to prevent the condensation of liquid vapours.

The mixing section contains a total of five mass flow controllers (MFC 2-6). The hydrocarbons are introduced into the mixing section via two mass flow controllers (MFC4 and MFC5) with an operation range of 0-25 ml min-1 (STP, helium based). The two

89


hydrocarbons can be mixed with a helium stream from either MFC3 (0-250 ml min-1 (STP)) in the manifold (MAN1) or MFC2 (0-5 ml min-1 (STP)). The smaller range of MFC2 allows a more stable and accurate operation at the lower helium flows (and higher partial pressure of the hydrocarbons). Although not used in this study, valve V1 allows the gas mixture to be directed to BPC1, where the mixture flow to the breakthrough column can be controlled with MFC6. In that situation a much lower mol fraction of the hydrocarbons can be created by further diluting the flow of this mixture with the helium stream from MFC2.

The desorption flow is controlled with a single mass flow controller (MFC1) with an operating range of 0-250 ml min-1 (STP). All the mass flow controllers are connected to the setup using VCR-couplings, which allows the easy exchange of mass flow controllers with each other at the various positions. This way, the operating range of the mass flow controllers in the setup can be easily changed.

The gas mixture first enters the convection oven (temperature range 298-338 K). In the oven the gas can be mixed with an evaporating liquid, which can be introduced by a piston pump (not used in this study). After passing the selection valve V3, the final gas mixture is fed to the top of the breakthrough column. Switching valve V3 either starts the loading or the desorption of the breakthrough column. For a breakthrough experiment the valve will be switched from helium to the gas mixture, while for a desorption experiment the valve will be returned to the other position.

The breakthrough column is installed inside the ceramic oven, which is located inside the convection oven. The maximum possible length of the column is 125 mm with an external diameter of approximately 50 mm. The pressure drop across the column is monitored with the differential pressure sensor (0-670 kPa). The ceramic oven allows the analysis of breakthrough and desorption profiles for temperatures up to 773 K. This large temperature range allows (part of) the preparation and pre-treatment of the adsorbents to be performed inside the breakthrough column in an inert atmosphere, without the chance of the exposure of the adsorbent to ambient air.

After the installation of the column and the pre-treatment, breakthrough and desorption experiments can be performed at different temperatures and various gas phase compositions. The pressure at the bottom of the breakthrough column is regulated with a back pressure controller (BPC 3). A slight pressure difference (8 kPa) between the analysis section and the breakthrough column results in a gas flow through the sample line and sample loops. This flow can be adjusted with the needle valve (V5). Selection valve V4 allows either the feed or the flow at the outlet of the breakthrough column to be sent to the gas chromatograph for analysis.

The operation of the breakthrough setup is automatically controlled with a computer on which Labview is installed. Once the breakthrough column is installed, a whole series of multiple breakthrough and desorption experiments at different flows, compositions, temperature, pressure and start times can be initiated. Continuous operations of more than 10 days have currently been performed without user intervention. Longer operations should also be possible.

90

Chapter 5

P

MFC

3C

V3PC

1SV

1Fi

lt1SR

V1

P

Filt2

MFC

4C

V4PC

2SV

2SR

V2

P

Filt3

MFC

5C

V5PC

3SV

3SR

V3

MA

N1

MFC

2C

V2

MFC

1C

V1

V1

MFC

6C

V6

V2

V3

SRV5

SRV4

Liqu

idpu

mp

FIA1

Cer

amic

ove

n

Bre

akth

roug

h co

lum

n

P

TCTS

V4

V5

BPC

1

Con

d1

P

BPC

2

Con

d2

P

BPC

3

P

Con

d3

FI2

Gas

chr

omat

ogra

ph

FID

s

Hel

ium

Com

pone

nt 2

Com

pone

nt 1

Ven

t

TCTS

Con

vect

ion

oven

P

MFC

3C

V3PC

1SV

1Fi

lt1SR

V1

P

Filt2

MFC

4C

V4PC

2SV

2SR

V2

P

Filt3

MFC

5C

V5PC

3SV

3SR

V3

MA

N1

MFC

2C

V2

MFC

1C

V1

V1

MFC

6C

V6

V2

V3

SRV5

SRV4

Liqu

idpu

mp

FIA1

Cer

amic

ove

n

Bre

akth

roug

h co

lum

n

P

TCTS

V4

V5

BPC

1

Con

d1

PP

BPC

2

Con

d2

PP

BPC

3

PP

Con

d3

FI2

Gas

chr

omat

ogra

ph

FID

s

Hel

ium

Com

pone

nt 2

Com

pone

nt 1

Ven

t

TCTS

Con

vect

ion

oven

Fig.

5.1

: Flo

w sh

eet o

f the

bre

akth

roug

h se

tup.

91


5.3.2 Analysis In order to determine the breakthrough and desorption profiles a fast analysis of the

mixture composition at the exit of the column is required. For single component mixtures, the concentrations can easily be monitored continuously and fast with for instance a mass spectrometer, an FID or TCD. However for binary hydrocarbon mixtures, and especially isomers, the analysis becomes more complex. Only when intense non-overlapping characteristic m/e peaks are present for each component, the composition can be determined by mass spectroscopy. For ethylene/ethane or propylene/propane mixtures most of the intense m/e peaks do overlap, but estimates of the individual concentrations can still be calculated based on the pure component fragmentation spectra. However reliable concentrations can only be obtained when the mol fractions of both components are of a similar order of magnitude. Otherwise the mass spectrum of one component will be completely hidden under the mass spectrum of the other component. In order to obtain more reliable concentration data various combinatorial techniques have been developed (Hagemeyer et al. (2001); Senkan et al. (2003)). These techniques either use multiple sensors or more complex equipment to obtain a better estimate of the mol fraction of both components.

The concentration of the individual components can much easier be determined with gas chromatography (GC), even over a wider concentration range. However, most of the conventional gas chromatographs require (a) minute(s) to analyze a single sample. To capture the whole breakthrough or desorption profile, this analysis time will be too long. One possibility for a faster sampling time is the temporary storage of the gas samples within the sample loops of multi-position valves. Unfortunately only a limited amount of samples can be stored and at the end of the experiment these samples still have to analyzed, which can take several hours. During this analysis no new breakthrough or desorption experiment can be started and the setup will be idle. In case the breakthrough time of the components is unknown, first a guess of the limited number of sampling times has to be made and better estimates of interesting sampling times can only be obtained after the analysis of the first breakthrough profile. To capture the most interesting part of the breakthrough profile a sequential process of adjusting the sampling times is required. This process of performing multiple breakthrough and desorption experiments results in a very long experimental procedure to capture only one breakthrough profile.

The continuous miniaturization of electronics and detectors and the continuous improvement of gas chromatography columns have resulted in the development of various designs of, so called, microGC’s. Compared to conventional gas chromatography, smaller tubing, valves, sample loops, columns, detectors and/or column ovens are used in these microGC’s. This results in a large reduction of the analysis time.

In our setup the CompactGC of Interscience is used to determine the mol fraction of both components. The GC is equipped with three parallel 8 meter Rt-QPlot capillary columns (diameter 0.32 mm) and each column is equipped with its own Flame Ionization Detector (FID). Whenever the GC is triggered, 45 samples can be analyzed continuously, after which 1-1½ minute is required to save the chromatogram on the computer. As an example, part of the chromatogram of a single GC-column for an ethylene/ethane mixture and a

92

Chapter 5

propylene/propane mixture is shown in Fig. 5.2a-b, respectively. These figures show the resulting gas chromatogram of 5 samples. Each sample results in two peaks, the first peak corresponds with the olefin, while the second peak corresponds with the paraffin. The total analysis/retention time of an ethylene/ethane sample is slightly less than 24 seconds. Therefore a sampling time of 24 seconds can be easily achieved for each GC-column. Since the GC is equipped with three similar GC-columns and detectors in parallel, this results in an effective sampling time of only 8 seconds. For propylene/propane mixtures the separation in the GC-column is more difficult. The total analysis/retention time is increased to approximately 65 seconds, although a further optimization may be possible. This would imply a significantly longer sampling time for propylene/propane mixtures. Fortunately, as seen in Fig. 5.2b the two peaks of the two components still fit within a 24 second time period. Therefore the consecutive injection of samples every 24 seconds will still allow a faster analysis. The front of two samples will then be travelling behind each other in the GC-column by 24 seconds apart. This approach allows a sampling time of 24 seconds for each GC-column for a propylene/propane mixture. With three parallel columns this results in a sampling time of 8 seconds for propylene/propane mixtures.

With a sampling time of 8 seconds and the possibility to take 45 samples, the CompactGC allows the continuous measurement of the mol fraction at the exit of the breakthrough column for 6 minutes. Thereafter a gap of 1-1½ minute is required to allow the detection of the last sample, to save the results and to reset the trigger of the EZChrom Elite software for the next chromatogram. Then again 45 samples can be analyzed during 6 minutes, followed by a 1-1½ minute gap, etc.

Time [min]

0.0 0.5 1.0 1.5 2.0

Sig

nal[

mV]

0.0

0.5

1.0

1.5C2H4

C2H6

C2H4

C2H6

C2H4

C2H6

C2H4

C2H6

C2H4

C2H6

a

Time [min]

0.0 0.5 1.0 1.5 2.0

Sig

nal[

mV]

0.0

0.5

1.0

1.5C2H4

C2H6

C2H4

C2H6

C2H4

C2H6

C2H4

C2H6

C2H4

C2H6

Time [min]

0.0 0.5 1.0 1.5 2.0

Sig

nal[

mV]

0.0

0.5

1.0

1.5C2H4

C2H6

C2H4

C2H6

C2H4

C2H6

C2H4

C2H6

C2H4

C2H6

a

0.0 0.5 1.0 1.5 2.00.0

0.2

0.4

0.6

Time [min]

Sign

al[m

V]

C3H6

C3H8

C3H6

C3H8

C3H6

C3H8

C3H6

C3H8

C3H6

C3H8

b

0.0 0.5 1.0 1.5 2.00.0

0.2

0.4

0.6

Time [min]

Sign

al[m

V]

C3H6

C3H8

C3H6

C3H8

C3H6

C3H8

C3H6

C3H8

C3H6

C3H8

0.0 0.5 1.0 1.5 2.00.0

0.2

0.4

0.6

Time [min]

Sign

al[m

V]

C3H6

C3H8

C3H6

C3H8

C3H6

C3H8

C3H6

C3H8

C3H6

C3H8

b

Fig. 5.2: Gas chromatogram of 5 samples for a) ethylene/ethane and b) propylene/propane.

93


The conversion of the peak areas in the gas chromatogram to the mol fraction of both components requires a calibration curve. As an example a calibration curve of one column for a propylene/propane mixture in helium is plotted in figure 5.3. The 15 points recorded within one chromatogram overlap perfectly, which indicates a good reproducibility within one chromatogram and indicates that the consecutive injection of multiple samples travelling behind each other in the GC-column did not influence the analysis. The calibration can be described by a straight line, which intersects the axes at the origin. Because of this linearity of the calibration line and its crossing at the origin, the mol fractions of both components can simply be calculated based on the peak areas obtained at the end of a breakthrough experiment, representing the feed composition at equilibrium, or before the beginning of a desorption experiment.

0 0.1 0.2 0.3 0.4 0.5Mol fraction [-]

0

2.5

5.0

7.5

Are

a[1

05A

.U.]

C3H6

C3H8

0 0.1 0.2 0.3 0.4 0.5Mol fraction [-]

0

2.5

5.0

7.5

Are

a[1

05A

.U.]

0 0.1 0.2 0.3 0.4 0.5Mol fraction [-]

0

2.5

5.0

7.5

Are

a[1

05A

.U.]

C3H6

C3H8

Fig. 5.3: Calibration lines of one GC-column for propane ( ) and propylene ( ).

A continuous analysis of a 50:50 binary mixture for 3 days showed that within an 8 h time period the peak area varied less than 1-2%. This stable value of the peak area, and the fact that each breakthrough or desorption experiment required less than 4½ hour, justifies the approach to calculate the mol fractions of both components from the equilibrium peak areas obtained at the end of the breakthrough experiment. This method to calculate the mol fractions was applied for all the experiments that were performed in this study. Once in a while a calibration line was recorded to check its linearity and the stability of the GC columns, but no large deviations were observed in this study.

5.3.3 Breakthrough column fillings For the breakthrough experiments multiple columns were used (Tables 5.2 and 5.3). The

SiC-column was packed with a sieve fraction of 63-71 m SiC particles (from a 150 mesh stock supply). The NaX crystals were synthesized following the recipe reported earlier (Van Miltenburg et al. (2006); Chapter 2 of this thesis). A sieve fraction of 63-71 m was used for the NaX-columns. A mixture of this sieve fraction and 36 wt% CuCl (Fluka, wt% is based on the dry sample mass of NaX) was subjected to heat treatment at 623 K in the small 36 wt% CuCl/NaX columns. For the large 36 wt% CuCl/NaX column a sieve fraction of larger (~100-120 m) NaX crystals, from multiple earlier syntheses, was used.

94

Chapter 5

The dimensions of the breakthrough columns are summarized in Table 5.2. The particles were contained inside the ¼”-tube using a stainless steel frit of 1 mm thick, with pore openings of 0.5 m, fixed in standard Swagelok couplings. For the 3/8

”-tube, the crystals were contained in the tube by quartz wool.

Table 5.2: Dimensions of the breakthrough columns. SiC NaX 36 wt% CuCl/NaX

Small 36 wt% CuCl/NaX

Large*

Outer diameter [inch] ¼ ¼ ¼ 3/8

Inner diameter [mm] 4.57 4.57 4.57 7.75 Length [mm] 60 60 60 125 Sample mass [g] 1.450 1.067** 1.286** 6.528**

Particle diameter [ m] 63-71 63-71 63-71 ~100-120 Ptot [kPa] 108 108 108 108

* Only used for flows and compositions of runs K and L (Table 5.4) at 408 K. ** The NaX crystals present in these samples contain 23 wt% water adsorbed from the ambient air, so the total dry mass will be lower.

The 36 wt% CuCl/NaX adsorbents were prepared following the optimized recipe reported earlier (Van Miltenburg et al. (2006); Chapter 2 of this thesis). Therefore a physical mixture of 36 wt% CuCl and NaX was put in the column tube and installed in the ceramic oven of the breakthrough setup. To prevent the contamination of the entire setup with subliming CuCl, the outlet of the adsorption column was first connected directly to a separate vent line. In a helium flow of 100 ml min-1 (SATP), the column was heated to 423 K at 1 K min-1, where it remained for 1 hour to remove most of the adsorbed water. Thereafter the temperature was further increased to 623 K at 1 K min-1. After 4 hours the temperature was decreased to room temperature. At room temperature the outlet of the column was disconnected from the separate vent line and immediately reconnected to the normal connection of the setup. To limit the exposure of the adsorbent to the atmosphere the helium flow was maintained, while the outlet connection was changed.

For the other columns, containing either NaX or SiC, the outlet was immediately connected to the normal connection of the setup. These columns were subjected to the same pre-treatment procedure. During the pre-treatment of these materials only adsorbed water desorbed from the particles and therefore contamination of the setup was not expected to occur.

5.3.4 Breakthrough and desorption experiments As shown in Table 5.2 four different columns were used for the breakthrough and

desorption experiments. Since the investigated gases do not adsorb on SiC, this column was used to investigate the breakthrough and desorption profile for the void space of the breakthrough column and to account for a possible time delay between the switch of valve V3 and the beginning of the detection of the breakthrough and desorption by the gas chromatograph.

95


The stability of the adsorbents and the reproducibility of the breakthrough profiles were tested earlier using a similar column as used for the other experiments in this study. The specification of the columns and the flow, pressure and temperature settings for these experiments can be found in Table 5.3. The breakthrough of an ethylene and ethane mixture on the 36 wt% CuCl/NaX adsorbent was recorded 3 times. For the NaX column the propylene and propane breakthrough profiles were recorded 8 times.

Table 5.3: Dimensions and settings for the stability and reproducibility tests. NaX 36 wt% CuCl/NaX

Outer diameter [inch] ¼ ¼Inner diameter [mm] 4.57 4.57 Length [mm] 60 60Sample mass [g] 1.039** 1.321**

Particle diameter [ m] 63-71 63-71 v, in, adsorption [ml SATP* min-1] 4 8

v, mol, adsorption [ mol min-1] 3.0 6.0 xparaffin [-] 0.25 0.25 xolefin [-] 0.25 0.25 T [K] 358 358 Ptot [kPa] 200 200

* SATP = Standard Ambient Temperature and Pressure (298 K and 101 kPa) ** The NaX crystals present in these samples contain 23 wt% water adsorbed from the ambient air.

Table 5.4: Total flows and compositions for the breakthrough and desorption experiments with the columns presented in Table 5.2.

Run tstart

[s]

v, in, adsorption

[ml SATP* min-1]

mol, in,

adsorption

[ mol s-1]

xparaffin, in

[-]

xolefin, in

[-]

v, in, desorption

[ml SATP* min-1]

mol, in,

desorption

[ mol s-1]A 250 2 1.5 0.000 1.000 20 15B 50 2 1.5 0.000 1.000 20 15C 250 80 60 0.025 0.025 20 15D 50 80 60 0.025 0.025 20 15E 250 40 30 0.050 0.050 20 15F 50 40 30 0.050 0.050 20 15G 250 160 120 0.0125 0.0125 20 15H 50 160 120 0.0125 0.0125 20 15I 250 254 189 0.00788 0.00788 20 15J 50 254 189 0.00788 0.00788 20 15K 250 8 6.0 0.250 0.250 20 15L 50 8 6.0 0.250 0.250 20 15M 250 4 3.0 0.500 0.500 20 15N 50 4 3.0 0.500 0.500 20 15O 250 2 1.5 1.000 0.000 20 15P 50 2 1.5 1.000 0.000 20 15

* SATP = Standard Ambient Temperature and Pressure (298 K and 101 kPa)

96

Chapter 5

The total flowrates and compositions used in the other breakthrough and desorption experiments are shown in Table 5.4. For these experiments the exit pressure of the breakthrough column was set at 108 kPa. After each desorption experiment the breakthrough column was flushed for 6 hours with helium at 125 ml min-1 (SATP) at the measurement temperature. Experiments showed that a temperature increase after these 6 hours did not result in the desorption of additional adsorbed hydrocarbons, which confirms that all gases have desorbed from the column at that time. The SiC, NaX and the small 36 wt% CuCl/NaX column were investigated for the whole series of flows and compositions shown in Table 5.4 at three temperatures (318, 358 and 408 K). For the NaX-column and the small 36 wt% CuCl/NaX-column, the breakthrough and desorption profile of (mixtures of) ethylene and ethane and (mixtures of) propylene and propane were investigated. On the SiC-column only the breakthrough and desorption of (mixtures of) propylene and propane were investigated. The results for propylene and propane mixtures on the SiC-column were assumed to be equal for the ethylene and ethane mixtures, since the adsorption on SiC was assumed to be zero for all the hydrocarbons. The large 36 wt% CuCl/NaX-column was only used for two experiments at 408 K for an ethylene, ethane and helium (25:25:50) mixture (runs K and L of Table 5.4).

The total analysis time of each breakthrough experiment was approximately 1 hour. Since the breakthrough of both components occurred much earlier, this implies an equilibration time of at least ½ hour. The desorption of the breakthrough column was performed isothermally with a helium flush of 20 ml min-1 (SATP) and was analyzed for 4 hours.

C3H6

C3H8

NaX, 408 K,Run K

0

0.25

0.50

Mol

frac

tion

[-]

0 500 1000 1500Time [s]

0

0.25

0.50

Mol

frac

tion

[-]

C3H6

C3H8

NaX, 408 K,Run L

C3H6

C3H8

NaX, 408 K,Run K

0

0.25

0.50

Mol

frac

tion

[-]

0 500 1000 1500Time [s]

0

0.25

0.50

Mol

frac

tion

[-]

C3H6

C3H8

NaX, 408 K,Run L

Fig. 5.4: Breakthrough profiles of propane ( ) and propylene ( ) mixture in helium (25:25:50, runs K & L) over NaX at 408 K. The time gaps in the GC-analysis are indicated in by the grey bars.

97


As shown in Table 5.4 and Fig. 5.4, two analysis start times (tstart) in the breakthrough and desorption experiments were used for two consecutive runs at similar temperature, flow and gas phase composition. As was mentioned earlier in the description of the analysis equipment of the setup, every 6 minutes a time gap of 1-1½ minute (the grey bars in Fig. 5.4) would be required to save and restart the analysis of the EZChrom Elite software. Because of the different start times, this gap will be located at a different position for the second run, which allows the capture of the complete breakthrough and desorption profile. The second run would then cover the time period where the gap was located in the first experiment.

5.3.5 Adsorptives Gases were provided by HoekLoos and had following purities: ethylene 2.8 (99.8%),

ethane 3.0 (99.9%), propylene 3.5 (99.95%), propane 3.5 (99.95%) and helium 4.6 (99.996%). The GC analysis of the olefins and paraffins indicated the presence of small quantities of the corresponding paraffin or olefin (e.g. ethane impurity in ethylene).

5.4 Results The reproducibility and stability of the eight breakthrough experiments on NaX and the

three experiments on the 36 wt% CuCl/NaX are shown in Fig. 5.5a-b. All the profiles overlap with each other. The breakthrough times for both components in Fig. 5.5a are about twice as long as those obtained later with a two times higher flowrate (runs K and L in Table 5.4) at 358 K and similar partial pressures. The breakthrough times for the 36 wt% CuCl/NaX column are approximately equal to those obtained later at a lower total pressure but at the same partial pressures and hydrocarbon flowrate (runs M and N in Table 5.4 at 358 K).

0 500 1000 1500 2000 2500 3000Time [s]

0

0.25

0.50

0.75

1.00

Mol

frac

tion

[-]

C3H6

C3H8

aNaX

0 500 1000 1500 2000 2500 3000Time [s]

0

0.25

0.50

0.75

1.00

Mol

frac

tion

[-]

C3H6

C3H8

0 500 1000 1500 2000 2500 3000Time [s]

0

0.25

0.50

0.75

1.00

Mol

frac

tion

[-]

C3H6

C3H8

aNaX

0 500 1000 1500 2000Time [s]

0

0.25

0.50

0.75

1.00

Mol

frac

tion

[-]

C2H4

C2H6b36 wt% CuCl/NaX

0 500 1000 1500 2000Time [s]

0

0.25

0.50

0.75

1.00

Mol

frac

tion

[-]

C2H4

C2H6

0 500 1000 1500 2000Time [s]

0

0.25

0.50

0.75

1.00

Mol

frac

tion

[-]

C2H4

C2H6b36 wt% CuCl/NaX

Fig. 5.5: Reproducibility tests; a) Breakthrough profiles of propane ( ) and propylene ( ) for 8 similar binary breakthrough experiments over NaX. b) Breakthrough profiles of ethane ( ) and ethylene ( ) for 3 similarbinary breakthrough experiments over 36 wt% CuCl/NaX. Conditions are in Table 5.3.

5.4.1 Calculation procedure As an example the procedure to calculate the adsorbed amounts of the olefins and paraffins

for the breakthrough and desorption experiments is explained in detail for one experiment. The breakthrough profile of a binary (50:50) mixture of ethylene and ethane in helium on the

98

Chapter 5

small 36 wt% CuCl/NaX-column at 408 K is presented in Fig. 5.6. For this breakthrough experiment the mol fractions of ethylene and ethane in the feed were set at 0.25 and the total feed flow was equal to 8 ml min-1 (SATP) (runs K and L of Table 5.4).

The breakthrough profile in this figure can be divided into 4 segments. In the first segment both components are adsorbed on the adsorbent. Since helium is present as a carrier gas, the sample lines of the GC are flushed and no hydrocarbons are detected.

At the end of the first segment the least adsorbing component (ethane in this case) starts to break through. As seen in Fig. 5.6 the mol fraction increases beyond the feed mol fraction of 0.25. This is simply the result of the absence of ethylene in the flow at the outlet of the column. The absence of the ethylene causes the mol fraction of ethane to increase to 1/3.During the second segment of the breakthrough profile only ethane in helium is observed in the gas chromatograph.

During the third segment of the breakthrough experiment, after about 6 minutes, ethylene starts to break through. The mol fraction of ethylene increases to its feed composition. The increase of the ethylene mol fraction causes the mol fraction of ethane to return to its feed composition.

Finally in the last segment of the breakthrough profile, the mol fractions of both components are equal to those in the feed. At that time equilibrium between the gas phase and the adsorbent is assumed to have been established. The average areas obtained in the chromatogram at this point, were used to determine the linear calibration lines and to calculate the molar compositions for each analysis point.

Under the assumption that helium does not adsorb in the column, the outlet flow of helium remains constant and is equal to its inlet flow. Based on the breakthrough profile and the known inlet flow and composition, Eq. 5.4 can be applied to calculate the outlet flow of each component during the breakthrough experiment. The result of this calculation is shown in Fig. 5.7. Once ethane has completely broken through, the outlet flow of ethane is equal to its inlet flow of 2 ml min-1 (SATP) and no roll-up is observed. Since the helium flow remains constant

0 250 500 750 1000Time [s]

0

0.25

0.50

Mol

arfra

ctio

n[-]

I II III IV

C2H6

C2H4

36 wt% CuCl/NaX

0 250 500 750 1000Time [s]

0

0.25

0.50

Mol

arfra

ctio

n[-]

I II III IV

C2H6

C2H4

0 250 500 750 1000Time [s]

0

0.25

0.50

Mol

arfra

ctio

n[-]

I II III IV

C2H6

C2H4

36 wt% CuCl/NaX

Fig. 5.6: Binary breakthrough profiles of ethane ( ) and ethylene ( )mixture in helium (1:1:2, runs K & L) over 36wt% CuCl/NaX at 408 K.

99


at 4 ml min-1 (SATP) and the ethylene flow out of the column is still zero, the mol fraction of ethane is then indeed equal to:

31

,,,,

,,

,,,,

,,, /

422

inheliumvinethanev

inethanev

outheliumvoutethanev

outethanevoutethanex (5.8)

Half an hour after the equilibrium was assumed to be established, desorption was started. The desorption profile of this loaded small 36 wt% CuCl/NaX is show in Fig. 5.8 in a semi logarithmic plot. Inserted is a detailed view of the first 100 seconds of the desorption profile. In the detailed view, the mol fractions of ethylene and ethane initially remain constant at the original equilibrium value that was reached at the end of the breakthrough experiment. This constant value is not associated with the adsorption process, but with the equipment and can be explained as follows. At first the remaining gas in the supply lines after valve V3, the void space in the column and the sample lines has to be replaced before the desorption of the adsorbent can be observed by the GC. After 30 seconds the ethane mol fraction quickly decreases to a very low value, while the mol fraction of ethylene decreases much slower.

0Time [s]

Mol

frac

tion

[-]

1000 1500 3000

10-1

1

10-2

10-3

10-4

500 25002000

C2H6

C2H4

0 40 60

10-1

10-2

10-3

10-4

20 80

C2H6

C2H436 wt% CuCl/NaX

0Time [s]

Mol

frac

tion

[-]

1000 1500 3000

10-1

1

10-2

10-3

10-4

500 25002000

C2H6

C2H4

0Time [s]

Mol

frac

tion

[-]

1000 1500 3000

10-1

1

10-2

10-3

10-4

500 25002000

C2H6

C2H4

0 40 60

10-1

10-2

10-3

10-4

20 80

C2H6

C2H4

0 40 60

10-1

10-2

10-3

10-4

20 80

C2H6

C2H436 wt% CuCl/NaX

Fig. 5.8: Desorption profiles of ethane ( ) and ethylene ( ) after the breakthrough presented in Fig. 5.5 and the binary equilibrium was established.

0 250 500 750 1000Time [s]

0

6

8Fl

owou

t [m

l (SA

TP) m

in-1

]

C2H6

C2H4

2

4He

Total

36 wt% CuCl/NaX

0 250 500 750 1000Time [s]

0

6

8Fl

owou

t [m

l (SA

TP) m

in-1

]

C2H6

C2H4

2

4He

Total

0 250 500 750 1000Time [s]

0

6

8Fl

owou

t [m

l (SA

TP) m

in-1

]

C2H6

C2H4

2

4He

Total

36 wt% CuCl/NaX

Fig. 5.7: Calculated outlet flows of ethane ( ), ethylene ( ) and helium (and the combined total flow ( ) for the breakthrough profile shown in Fig. 5.5.

100

Chapter 5

In agreement with this interpretation, the ‘desorption’ profile of the SiC-column presented in Fig. 5.14c-d shows a similar delay. For 30 seconds the mol fractions of both components remain constant and are equal to the feed/equilibrium values, after which they rapidly decrease to lower values. To account for this delay, the first 30 seconds of the desorption profiles were disregarded for further calculations of the adsorbed amounts.

These results shows that the use of Eq. 5.6 has some limitations. During the first 30 seconds the composition remains almost equal to the feed composition at these conditions, and therefore the application of Eq. 5.6 could yield unrealistic high (or even negative) outlet flows for the breakthrough experiment. Only after the initial 30 seconds, Eq. 5.6 is valid and can be used to calculate the desorption flows out of the breakthrough column. The outcome of this calculation is shown in Fig. 5.9, for the desorption profiles presented in Fig. 5.8.

After a certain desorption time the mol fractions of both components in the desorption profiles (Fig. 5.8) are below the detection limit of the GC-analysis. Sometimes, especially for the olefins at the lowest temperature, the desorption was not yet completed after the analysis time of 4 hours. In order to estimate the amount that still remained on the breakthrough column, the last part of the curve was fitted to an exponentially decaying function and interpreted analytically.

In Fig. 5.10 the integral of the amount that desorbed from this loaded breakthrough column (i.e. the result of Eq. 5.7) is plotted versus the desorption time. As seen in this graph, the amount of ethane desorbing from the column is very small. For ethylene a fast initial desorption is observed and then its molar fraction decreases asymptotically to zero. In order to calculate the total amount desorbed from the column, the last part was approximated with the integral of the exponential decay function that was fitted on the last points of Fig. 5.8. Fortunately, the 4 h analysis period allowed all desorption profiles to be analyzed until very low mol fractions. Therefore the integral only yielded an additional increase in the amount desorbed from the column of at maximum 1-2% of the total loading, so representing only a small adjustment.

0Time [s]

Flow

out [

ml (

SA

TP) m

in-1

]

1000 1500 3000

1

10

10-1

10-2

10-3

500 25002000

C2H6

C2H4

He

36 wt% CuCl/NaX

0Time [s]

Flow

out [

ml (

SA

TP) m

in-1

]

1000 1500 3000

1

10

10-1

10-2

10-3

500 25002000

C2H6

C2H4

He

0Time [s]

Flow

out [

ml (

SA

TP) m

in-1

]

1000 1500 3000

1

10

10-1

10-2

10-3

500 25002000

C2H6

C2H4

He

36 wt% CuCl/NaX

Fig. 5.9: Calculated outlet flow during the desorption for ethane ( ), ethylene ( ) and helium ( for the desorption profiles shown in Fig. 5.7.

101


0 500 1000 1500 2000 2500 3000Time [s]

0

0.25

0.50

0.75

1.00

Load

ing

[mol

kg-1

]

C2H4

C2H6

36 wt% CuCl/NaX

0 500 1000 1500 2000 2500 3000Time [s]

0

0.25

0.50

0.75

1.00

Load

ing

[mol

kg-1

]

C2H4

C2H6

0 500 1000 1500 2000 2500 3000Time [s]

0

0.25

0.50

0.75

1.00

Load

ing

[mol

kg-1

]

C2H4

C2H6

36 wt% CuCl/NaX

Fig. 5.10: Calculated integral amount of ethane ( ) and ethylene ( )desorbed from the column, based on the desorption profile shown in Fig. 5.7.

5.4.2 36 wt% CuCl/NaX columns The breakthrough profiles of the ethylene and ethane mixture at 318 and 358 K are

presented in Fig. 5.11a-b. The breakthrough profiles at 408 K was shown earlier in Fig. 5.6. At 358 K the breakthrough time of ethylene is longer than at 408 K. The length of the third segment in the breakthrough profile of ethylene, this is the segment in which the binary equilibrium is established, has also increased. Initially the mol fraction of ethylene increased rapidly to about 75% of the feed composition, after which it slowly approached the feed mol fraction of 0.25. The breakthrough time of ethane is similar at 358 and 408 K, which is apparently hardly affected by the temperature at these conditions.

At 318 K the breakthrough of ethylene occurs much earlier than at the other two temperatures. Furthermore, the time to reach the binary equilibrium is considerably extended (segment III in the figures). After 150 seconds the ethylene mol fraction increased quickly to about 75% of the feed composition, after which it very slowly approached a mol fraction of 0.25. Although very close, even after 40 minutes the mol fractions of both ethylene and

0 250 500 750 1000Time [s]

0

0.25

0.50

Mol

frac

tion

[-]

I II III

C2H4

C2H6

a36 wt% CuCl/NaX

0 250 500 750 1000Time [s]

0

0.25

0.50

Mol

frac

tion

[-]

I II III

C2H4

C2H6

0 250 500 750 1000Time [s]

0

0.25

0.50

Mol

frac

tion

[-]

I II III

C2H4

C2H6

a36 wt% CuCl/NaX

0 250 500 750 1000Time [s]

0

0.25

0.50

Mol

frac

tion

[-]

C2H4

C2H6

I II III

b36 wt% CuCl/NaX

0 250 500 750 1000Time [s]

0

0.25

0.50

Mol

frac

tion

[-]

C2H4

C2H6

I II III

0 250 500 750 1000Time [s]

0

0.25

0.50

Mol

frac

tion

[-]

C2H4

C2H6

I II III

b36 wt% CuCl/NaX

Fig. 5.11: Binary breakthrough profiles of ethane ( ) and ethylene ( ) mixture in helium (1:1:2, runs K & L) over 36wt% CuCl/NaX at a) 318 K and b) 358 K.

102

Chapter 5

ethane did not yet completely reach the feed composition and equilibrium values. Unfortunately, the small fluctuation/error in the experimental analysis with the GC does not allow an exact confirmation of the establishment of equilibrium.

At a lower mol fraction for both components and a higher total flow (runs I and J), the shape of the breakthrough profiles will change. In Fig. 5.12 the breakthrough profiles of a diluted binary (50:50) mixture of ethylene and ethane on the small 36 wt% CuCl/NaX-column at 408 K is shown. In order to achieve a mol fraction of 0.00788 the helium flow was increased from 4 to 250 ml min-1 (SATP), while the flows of ethylene and ethane were both maintained at 2 ml min-1 (SATP). This way breakthrough of both components would approximately occur within a similar time frame over the investigated pressure range (0.8-54 kPa) for a type I adsorption isotherm (Brunauer classification) with a large equilibrium constant, meaning a large adsorption affinity. Compared to Fig. 5.6 the breakthrough profiles of both ethylene and ethane are much broader. The time difference between the breakthrough profiles of ethylene and ethane is reduced considerably and the profiles have become less steep.

The effect of the increased flowrate on the dynamics of the breakthrough column can be determined on the SiC-column. In Fig. 5.14a-b the breakthrough profiles over the SiC-column for a total flow of 8 and 254 ml min-1 (SATP) are pictured. The increased flow broadens the breakthrough profiles of both components. Besides the effect on the dynamics, the breakthrough profiles over the SiC-column confirm this, they have indeed a similar shape for both components when these are not adsorbed on the column. Just as for the desorption profiles (Fig. 5.14c-d), the breakthrough profiles show an initial delay before the breakthrough of both components is detected with the GC, which is shorter at higher flowrates.

0 250 500 750 1000Time [s]

0

0.005

0.010

Mol

frac

tion

[-]

C2H6

C2H4

36 wt% CuCl/NaX

0 250 500 750 1000Time [s]

0

0.005

0.010

Mol

frac

tion

[-]

C2H6

C2H4

0 250 500 750 1000Time [s]

0

0.005

0.010

Mol

frac

tion

[-]

C2H6

C2H4

36 wt% CuCl/NaX

Fig. 5.12: Binary breakthrough profiles of ethane ( ) and ethylene ( )mixture in helium (1:1:125, runs I & J) over 36wt% CuCl/NaX at 408 K.

103


0 250 500 750 1000Time [s]

0

0.005

0.010

0.015

Mol

frac

tion

[-]C2H6

C2H4

36 wt% CuCl/NaX

0 250 500 750 1000Time [s]

0

0.005

0.010

0.015

Mol

frac

tion

[-]C2H6

C2H4

0 250 500 750 1000Time [s]

0

0.005

0.010

0.015

Mol

frac

tion

[-]C2H6

C2H4

36 wt% CuCl/NaX

Fig. 5.13: Desorption profiles of ethane ( ) and ethylene ( ) after the breakthrough presented in Fig. 5.11 and the binary equilibrium was established.

Desorption of the small 36 wt% CuCl/NaX-column loaded with a diluted ethylene and ethane mixture (mol fraction 0.00788) was started 1 h after the start of the breakthrough experiment. At that time the equilibration time was at least ½ h. The desorption profiles are presented in Fig. 5.13. In this figure the mol fraction of ethane decreased quickly to zero. For ethylene first an increase in the mol fraction is observed, after which it slowly decreased to

0 50 100 150 200 250Time [s]

0

0.25

0.50

Mol

frac

tion

[-]

a SiC

0 50 100 150 200 250Time [s]

0

0.25

0.50

Mol

frac

tion

[-]

0 50 100 150 200 250Time [s]

0

0.25

0.50

Mol

frac

tion

[-]

a SiC

0 250 500 750Time [s]

0

0.005

0.010M

ol fr

actio

n[-]

b SiC

0 250 500 750Time [s]

0

0.005

0.010M

ol fr

actio

n[-]

0 250 500 750Time [s]

0

0.005

0.010M

ol fr

actio

n[-]

b SiC

0 50 100 150 200 250Time [s]

0

0.1

0.3

Mol

frac

tion

[-] 0.2

cSiC

0 50 100 150 200 250Time [s]

0

0.1

0.3

Mol

frac

tion

[-] 0.2

0 50 100 150 200 250Time [s]

0

0.1

0.3

Mol

frac

tion

[-] 0.2

cSiC

0 50 100 150 200 250Time [s]

0

0.005

0.010

Mol

frac

tion

[-]

dSiC

0 50 100 150 200 250Time [s]

0

0.005

0.010

Mol

frac

tion

[-]

0 50 100 150 200 250Time [s]

0

0.005

0.010

Mol

frac

tion

[-]

dSiC

Fig. 5.14: Binary breakthrough (a & b) and ‘desorption’ (c & d) profiles of propane ( ) and propylene ( )mixtures over SiC for a low flow and high partial pressure (run K & L and figures a & c) or a high flow and low partial pressure (run I & J and figures b & d) at 358 K. Lines are to guide the eye.

104

Chapter 5

lower mol fractions. This initial increase in the desorption profiles of the olefins was also observed for the low mol fractions of 0.025 and 0.0125, whereby the total flow during the breakthrough experiment was set to 80 or 160 ml min-1 (SATP) respectively (runs C & D and G & H). This increase was however not observed for the SiC-column shown in Fig. 5.14d,which indicates that this increase is related to the adsorption/desorption process.

In Fig. 5.15a-c the breakthrough profiles of a binary (50:50) mixture of propylene and propane are shown at the three temperatures (runs K and L). Compared to ethane, the breakthrough of propane occurred at a later time at 358 and 408 K. At these temperatures the mol fraction of propane first increased rapidly to ~ 1/3, after which a roll-up of the mol fraction to 0.40 is seen. At that moment propylene started to break through and the mol fraction of propane returned to the feed fraction of 0.25.

At 408 K the breakthrough time for propylene was approximately equal to the breakthrough time for ethylene at similar conditions (Fig. 5.6). At lower temperatures the breakthrough of propylene occurred earlier. At these temperatures the mol fraction of propylene first increases quickly to about 75% of the equilibrium value, where after it slowly increases further to 0.25. At 318 K the breakthrough times of propylene and propane are approximately equal, although for a short period a roll-up of propane is seen. The fast

0 250 500 750 1000Time [s]

0

0.25

0.50

Mol

frac

tion

[-] C3H8

C3H6

a36 wt% CuCl/NaX

0 250 500 750 1000Time [s]

0

0.25

0.50

Mol

frac

tion

[-] C3H8

C3H6

0 250 500 750 1000Time [s]

0

0.25

0.50

Mol

frac

tion

[-] C3H8

C3H6

a36 wt% CuCl/NaX

0 250 500 750 1000Time [s]

0

0.25

0.50

Mol

frac

tion

[-]C3H8

C3H6

b36 wt% CuCl/NaX

0 250 500 750 1000Time [s]

0

0.25

0.50

Mol

frac

tion

[-]C3H8

C3H6

0 250 500 750 1000Time [s]

0

0.25

0.50

Mol

frac

tion

[-]C3H8

C3H6

b36 wt% CuCl/NaX

0 250 500 750 1000Time [s]

0

0.25

0.50

Mol

frac

tion

[-]

C3H8

C3H6

c36 wt% CuCl/NaX

0 250 500 750 1000Time [s]

0

0.25

0.50

Mol

frac

tion

[-]

C3H8

C3H6

0 250 500 750 1000Time [s]

0

0.25

0.50

Mol

frac

tion

[-]

C3H8

C3H6

c36 wt% CuCl/NaX

Fig. 5.15: Binary breakthrough profiles of propane ( ) and propylene ( ) mixture in helium (1:1:2, runs K & L) over 36wt% CuCl/NaX at a) 318 K, b) 358 K and c) 408 K.

105


breakthrough of propylene at this temperature resulted in the immediate decrease of the mol fraction of propane to the equilibrium value.

The breakthrough and the desorption profiles for the ethylene and ethane mixture on the larger 36 wt% CuCl/NaX column at 408 K are plotted in Fig. 5.16a-b. The breakthrough profiles show an early breakthrough of ethane after about 100 seconds and the mol fraction rises to 1/3. The ethane mol fraction remains at this value until ethylene starts to break through after ~ 1000 seconds. The mol fraction of ethylene rises very fast after which it slowly evolves to the equilibrium value. Compared to the small column the breakthrough times of the large column were a factor 2½ longer, whilst the large column was filled with approximately 5 times more adsorbent of a larger crystal size, and it was operated at the same temperature, pressure and flowrates.

The desorption profiles showed a fast decrease of the ethane mol fraction on the large 36 wt% CuCl/NaX column. Already after 250 seconds the mol fraction of ethane was smaller than 1% of the original equilibrium value, while the ethylene mol fraction had only decreased by 50% at that time. Only after 3000 seconds the mol fraction of ethylene becomes smaller than 1% of the original feed composition and desorption is almost completed.

Based on the desorption profiles and by applying Eq. 5.7 the loading of both components on the large column was calculated. In agreement with the breakthrough profile and Eq. 5.5, this yields an increase in the calculated total adsorbed amount by a factor 2½ for both components on the large column compared to the small column, whereas the total amount of adsorbent, of a larger crystal size, was increased by a factor 5.

0 500 1500 2000 2500Time [s]

0

0.25

0.50

Mol

frac

tion

[-]

C2H6

C2H4

1000

a36 wt% CuCl/NaX

0 500 1500 2000 2500Time [s]

0

0.25

0.50

Mol

frac

tion

[-]

C2H6

C2H4

10000 500 1500 2000 2500Time [s]

0

0.25

0.50

Mol

frac

tion

[-]

C2H6

C2H4

1000

a36 wt% CuCl/NaX

0Time [s]

Mol

frac

tion

[-]

2500 7500

10-1

1

10-2

10-3

10-4

5000

C2H6

C2H4

b36 wt% CuCl/NaX

0Time [s]

Mol

frac

tion

[-]

2500 7500

10-1

1

10-2

10-3

10-4

5000

C2H6

C2H4

0Time [s]

Mol

frac

tion

[-]

2500 7500

10-1

1

10-2

10-3

10-4

5000

C2H6

C2H4

b36 wt% CuCl/NaX

Fig. 5.16: Binary breakthrough (a) and desorption (b) profiles of ethane ( ) and ethylene ( ) mixtures in helium (1:1:2, runs K & L) over the large 36 wt% CuCl/NaX column at 408 K.

106

Chapter 5

5.4.3 NaX-column In Fig. 5.17a-b the breakthrough profiles of the binary mixture of ethylene and ethane for

the NaX-column are shown. The breakthrough of ethylene at 318 K occurs approximately 6 times later than for the small 36 wt% CuCl/NaX-column shown in 5.10a. Compared to that column, the mol fraction of ethylene quickly rises to the feed composition on the NaX column. On the small 36 wt% CuCl/NaX-column a considerable time was required to reach the final binary equilibrium, while this equilibrium appears to be reached more quickly on the NaX-column.

The breakthrough time of ethane was also considerably increased by approximately a factor 7 on the NaX-column. On the small 36 wt% CuCl/NaX-column the breakthrough of ethane occurs at approximately the same time as for the SiC-column, while for the NaX-column a much longer delay was observed. After the breakthrough of ethane on the NaX-column, a roll-up of ethane is seen and the mol fraction increases beyond the mol fraction of 1/3 that would have been expected on the basis of only the complete adsorption of ethylene from the feed (see Eq. 5.8). The later breakthrough and the roll-up of ethane indicate that besides ethylene also ethane is adsorbed at first.

0 500 1500 2000Time [s]

0

0.25

0.50

Mol

frac

tion

[-]

C2H6

C2H4

1000

aNaX

0 500 1500 2000Time [s]

0

0.25

0.50

Mol

frac

tion

[-]

C2H6

C2H4

10000 500 1500 2000Time [s]

0

0.25

0.50

Mol

frac

tion

[-]

C2H6

C2H4

1000

aNaX

0 100 300 400 500Time [s]

0

0.25

0.50

Mol

frac

tion

[-]

C2H6

C2H4

200

bNaX

0 100 300 400 500Time [s]

0

0.25

0.50

Mol

frac

tion

[-]

C2H6

C2H4

2000 100 300 400 500Time [s]

0

0.25

0.50

Mol

frac

tion

[-]

C2H6

C2H4

200

bNaX

Fig. 5.17: Binary breakthrough profiles of ethane ( ) and ethylene ( ) mixture in helium (1:1:2, runs K & L) over NaX at a) 318 K and b) 408 K.

At 408 K the breakthrough times of both ethane and ethylene are considerably reduced, viz., by a factor 5-6 compared to 318 K. The mol fraction of ethane still showed a roll-up beyond 1/3, but only for a relatively short period of time. It is interesting to compare this breakthrough profile at 408 K with the breakthrough profile on the small 36 wt% CuCl/NaX-column presented in Fig. 5.6. The dispersion of CuCl in the pores of the zeolite resulted in an increase in the breakthrough time of ethylene by a factor 2, while the breakthrough time of ethane was reduced. So the time delay between ethane and ethylene breakthrough for NaX is smaller than for the CuCl/NaX sample, suggesting a better separation over the later sample.

107


In Fig. 5.18 the desorption profiles of the NaX-column loaded with ethylene and ethane at 318 K are shown. The mol fraction of both components decreases gradually to lower values during the desorption. The mol fraction of ethylene shows an interruption in the decrease close to the time where ethane had completely desorbed from the column. At this interruption the mol fraction of ethylene remained almost constant for a short period of time, after which it continued to decrease further at a similar rate.

0 1000 2000 3000Time [s]

0

0.1

0.2

0.3

Mol

frac

tion

[-]C2H6

C2H4

NaX

0 1000 2000 3000Time [s]

0

0.1

0.2

0.3

Mol

frac

tion

[-]C2H6

C2H4

0 1000 2000 3000Time [s]

0

0.1

0.2

0.3

Mol

frac

tion

[-]C2H6

C2H4

NaX

Fig. 5.18: Desorption profiles of ethane ( ) and ethylene ( ) after the breakthrough presented in Fig. 5.16a and the binary equilibrium was established.

The breakthrough profiles of the propylene and propane mixture on NaX are shown in Fig. 5.19a-b. Just like ethane, a roll-up of the propane mol fraction beyond 1/3 is seen on the NaX-column. The breakthrough times of propylene and propane were considerably longer on the NaX-column compared to those observed on the small 36 wt% CuCl/NaX column shown in Fig. 5.15a-c. On that column only a minor time difference between the first breakthrough points of propylene and propane was observed at 318 K, whereas on the NaX-column the breakthrough time of propylene was almost two times longer as that of propane.

0 500 1000 1500Time [s]

0

0.25

0.50

Mol

frac

tion

[-]

C3H6

C3H8 aNaX

0 500 1000 1500Time [s]

0

0.25

0.50

Mol

frac

tion

[-]

C3H6

C3H8

0 500 1000 1500Time [s]

0

0.25

0.50

Mol

frac

tion

[-]

C3H6

C3H8 aNaX

0 500 1000 1500Time [s]

0

0.25

0.50

Mol

frac

tion

[-]

C3H6

C3H8

bNaX

0 500 1000 1500Time [s]

0

0.25

0.50

Mol

frac

tion

[-]

C3H6

C3H8

0 500 1000 1500Time [s]

0

0.25

0.50

Mol

frac

tion

[-]

C3H6

C3H8

bNaX

Fig. 5.19: Binary breakthrough profiles of propane ( ) and propylene ( ) mixture in helium (1:1:2, runs K & L) over NaX at a) 318 K and b) 408 K.

108

Chapter 5

At 408 K the breakthrough times of propylene and propane on NaX are reduced by 20% compared to those on NaX at 318 K. This reduction is much smaller than the reduction observed for the ethylene/ethane mixture. The breakthrough times of propylene and propane still occurred relatively late at the higher temperature. At 408 K the dispersion of CuCl in the pores of the zeolite resulted in a reduction of the breakthrough times for both propane and propylene by approximately a factor 2½ compared to NaX as can be concluded from a comparison of Fig. 5.19b with Fig. 5.15c.

5.4.4 Single component breakthrough In Fig. 5.20a-b the single component breakthrough profiles for ethylene and ethane on the

NaX-column at 358 K are presented. Because of the impurities in the ‘pure’ gas feeds, the breakthrough of two components is observed. These breakthrough profiles showed some interesting phenomena concerning the separation performance of the adsorbent.

The breakthrough profile of ethylene (Fig. 5.20a) initially showed a relatively high mol fraction of the ethane impurity present in the ‘pure’ ethylene feed. After the breakthrough of ethylene, the ethane mol fraction quickly decreases to the supply specification of the ethylene gas cylinder (99.8%). A similar increase of the paraffin mol fraction was observed at the other temperatures and for propylene. Because of the lower ethane capacity this increase was relatively small on the small 36 wt% CuCl/NaX column.

For ethane the breakthrough of the ethylene impurity in the ethane gas cylinder is seen in the profiles for the NaX-column at 358 K (Fig. 5.20b) and 408 K. For the lowest temperature (318 K) this breakthrough is not observed. Also for propane the breakthrough of the propylene impurity is not observed on the NaX-column during the time of the experiment. On the small 36wt% CuCl/NaX-column the breakthrough of both olefin impurities was not observed within the analysis time of 1 h for both the ethane and the propane gas cylinders. The exposure of the larger 36 wt% CuCl/NaX column to an ethane flow of 2 ml min-1 (SATP) didn’t show any breakthrough of ethylene for at least 5 days, as could have been expected based on the single component adsorption isotherm of ethylene and the amount of adsorbent present in the column. These results confirm the higher affinity of the CuCl/NaX adsorbent for the olefins.

1Time [s]

Mol

frac

tion

[-]

2000

10-1

1

10-2

10-3

10-4

1000 3000

C2H6

C2H4a NaX

1Time [s]

Mol

frac

tion

[-]

2000

10-1

1

10-2

10-3

10-4

1000 3000

C2H6

C2H4a

1Time [s]

Mol

frac

tion

[-]

2000

10-1

1

10-2

10-3

10-4

1000 3000

C2H6

C2H4a NaX

Time [s]1

Mol

frac

tion

[-]

1000 2000

10-1

1

10-2

10-3

10-4

3000

C2H6

C2H4

bNaX

Time [s]1

Mol

frac

tion

[-]

1000 2000

10-1

1

10-2

10-3

10-4

3000

C2H6

C2H4

b

Time [s]1

Mol

frac

tion

[-]

1000 2000

10-1

1

10-2

10-3

10-4

3000

C2H6

C2H4

bNaX

Fig. 5.20: Breakthrough profile of ethylene (a) and ethane (b) at 358 K over NaX. Ethylene ( ), ethane ( ).Lines are to guide the eye. Due to the adsorption process the impurities are clearly visibly, see text.

109


5.4.5 Adsorption capacities and selectivities Using Eq. 5.5 and Eq. 5.7 the adsorbed amounts for all the adsorption and desorption

experiments were calculated. The results of these calculations are shown in Fig. 5.21a-b for ethylene and ethane mixtures and Fig. 5.22a-b for propylene and propane mixtures on the 36 wt% CuCl/NaX-columns. The calculation of the adsorbed amounts using the breakthrough profiles (Eq. 5.5) instead of the desorption profiles (Eq. 5.7) in many cases resulted in a similar outcome as can be seen in the figures, though at the lower pressures, especially for the paraffins, the calculation based on the breakthrough profiles yielded higher amounts than those based on the desorption profiles.

At the end of the breakthrough profile the small absolute analytic error of the GC made it difficult to distinguish the paraffin mol fraction from the equilibrium value at these lower partial pressures. This resulted in a larger relative error in the integral adsorbed amounts of the paraffins based on the breakthrough profile at these conditions. This can be seen in Fig. 5.27, in which the relative experimental error in the adsorbed amount is shown for propane and propylene on 36 wt% CuCl/NaX at 358 K. Similar trends in the relative error were

10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1a

36 wt% CuCl/NaX

318 K

358 K

408 K

10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1a

36 wt% CuCl/NaX

318 K

358 K

408 K

10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1b

36 wt% CuCl/NaX

318 K

358 K

408 K

10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1b

36 wt% CuCl/NaX

318 K

358 K

408 K

10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1

318 K

358 K 408 K

c

36 wt% CuCl/NaX10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1

318 K

358 K 408 K

10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1

318 K

358 K 408 K

c

36 wt% CuCl/NaX10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1318 K

358 K

408 K

d

36 wt% CuCl/NaX10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1318 K

358 K

408 K

10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1318 K

358 K

408 K

d

36 wt% CuCl/NaX

Fig. 5.21: a-b) Loadings of (a) ethane and (b) ethylene for binary mixture (50:50) adsorption at 318 K ( , ), 358 K ( , and 408 K ( , ) on 36 wt% CuCl/NaX. (Open symbols based on breakthrough profile, closed symbols based on desorption profile.) Lines follow the closed symbols and are to guide the eye. c-d) IAS-theory prediction for the loading of (c) ethane and (d) ethylene for a (50:50) mixture on 36 wt% CuCl/NaX.

110

Chapter 5

observed for the other adsorptive-adsorbent combinations at the various temperatures. Because of these larger errors in the adsorbed amounts calculated from the breakthrough profiles, primarily the points obtained from the desorption profile were used for the lines to guide the eye. At the higher pressures the obtained loadings and trends from the breakthrough profiles were almost equal to those obtained from the desorption profiles, in agreement with the smaller experimental error at these pressures.

The adsorbed amounts for the olefins on 36 wt% CuCl/NaX are higher than those for the paraffins at all the investigated temperatures and pressures. Almost all components show an increase in the adsorbed amount when their partial pressure is increased, except at 318 K, where a decrease is observed at the higher pressures. An increase in the temperature shows different effects. At the lowest temperature (318 K) the adsorbed amounts of both components on the 36 wt% CuCl/NaX adsorbent are often lower than those determined at higher temperatures. This phenomenon is more pronounced for the propane and propylene. Whenever the temperature is increased, a decrease in the ethane uptake is seen at low pressures, while at high pressures the opposite is observed.

10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1a

36 wt% CuCl/NaX

318 K

358 K408 K

10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1a

36 wt% CuCl/NaX

318 K

358 K408 K

10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1b

36 wt% CuCl/NaX

318 K

358 K

408 K

10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1b

36 wt% CuCl/NaX

318 K

358 K

408 K

10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1

318 K

358 K

408 K

c

36 wt% CuCl/NaX10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1

318 K

358 K

408 K

10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1

318 K

358 K

408 K

c

36 wt% CuCl/NaX10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1318 K

358 K

408 K

d

36 wt% CuCl/NaX10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1318 K

358 K

408 K

10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1318 K

358 K

408 K

d

36 wt% CuCl/NaX

Fig. 5.22: a-b) Loadings of (a) propane and (b) propylene for binary mixture (50:50) adsorption at 318 K ( , ), 358 K ( , and 408 K ( , ) on 36 wt% CuCl/NaX. (Open symbols based on breakthrough profile, closed symbols based on desorption profile.) Lines follow the closed symbols and are to guide the eye. c-d) IAS-theory prediction for the loading of (c) propane and (d) propylene for a (50:50) mixture on 36 wt% CuCl/NaX.

111


The low pressure calculations using Eq. 5.5 for the breakthrough profiles yielded often smaller uptake values than for the non-adsorbing SiC-column. The small absolute error of the analysis close to the equilibrium resulted in a large error in the integral of the adsorbed amounts. In two experiments (runs K and L at 318 K for a C3-mixture and runs K and L at 408 K for a C2-mixture) also the outcome of Eq. 5.7 was smaller than the outcome for the non-adsorbing SiC-column. In all these cases, the integration and the subtraction of the SiC-column results in an unrealistic negative adsorbed amount. These negative values were not included in the graphs.

Based on the single component isotherms (Chapter 4 of this thesis) and the IAS-theory, the adsorbed amounts were estimated for a binary (50:50) mixture. The results of these calculations for the 36 wt% CuCl/NaX adsorbent is shown in Figs. 5.21c-d and 5.22c-d for binary mixtures of ethylene/ethane or propylene/propane, respectively. For the olefins this predicts an increase in the adsorbed amount when either the pressure is increased or the temperature is decreased. Analogously, for the paraffins an increase in temperature results in a decrease of the adsorbed amount at the lower partial pressures, but at the higher pressures an

10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1a

NaX

318 K

358 K

408 K

10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1a

NaX

318 K

358 K

408 K

10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1b

NaX

318 K

358 K408 K

10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1b

NaX

318 K

358 K408 K

10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1

318 K

358 K

408 K

c

NaX10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1

318 K

358 K

408 K

10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1

318 K

358 K

408 K

c

NaX10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1

318 K

358 K

408 K

d

NaX10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1

318 K

358 K

408 K

10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1

318 K

358 K

408 K

d

NaX

Fig. 5.23: a-b) Loadings of (a) ethane and (b) ethylene for binary mixture (50:50) adsorption at 318 K ( , ), 358 K ( , and 408 K ( , ) on NaX. (Open symbols based on breakthrough profile, closed symbols based on desorption profile.) Lines follow the closed symbols and are to guide the eye. c-d) IAS-theory prediction for the loading of (c) ethane and (d) ethylene for a (50:50) mixture on NaX.

112

Chapter 5

increase in the adsorbed amount is observed when the temperature is increased, which is in agreement with some of the experimental data.

The calculated amounts adsorbed on the NaX-column are shown in Figs. 5.23a-b and 5.24a-b. The adsorbed amounts on the NaX-column are larger than those on the small 36 wt% CuCl/NaX-column. An increase in the partial pressures results in an increase in the adsorbed amounts. A larger adsorbed amount is found for the olefins than for the paraffins. An increase in the temperature does result in a decrease of the adsorbed amounts for ethylene/ethane mixtures. For propylene/propane mixtures a temperature increase results in an increase in the adsorbed amount of the propane at higher partial pressures, while the reverse is observed at lower partial pressures.

The isotherms of the single components (Chapter 4 of this thesis) were used to calculate the binary adsorbed amounts of olefin/paraffin mixtures on the NaX-column with the IAS-theory. The results of these calculations are shown in Figs. 5.23c-d and 5.24c-d. The outcome of these calculations show large similarities with the experimental data, both qualitatively and quantitatively. The figures show an increase in the adsorbed amount of both olefins and

10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1a

NaX

318 K358 K

408 K

10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1a

NaX

318 K358 K

408 K

10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1b

NaX

318 K

358 K

408 K

10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1b

NaX

318 K

358 K

408 K

10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1

318 K

358 K

408 K

c

NaX10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1

318 K

358 K

408 K

c

NaX10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1 318 K

358 K

408 K

d

NaX10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1 318 K

358 K

408 K

10-4

Pressure [kPa]

q[m

ol k

g-1]

102

10

10110-1

10-3

10-2

10-1

1 318 K

358 K

408 K

d

NaX

Fig. 5.24: a-b) Loadings of (a) propane and (b) propylene for binary mixture (50:50) adsorption at 318 K ( , ), 358 K ( , and 408 K ( , ) on NaX. And the literature data ( ) for a binary propane and propylene mixture on NaX at 343 K (Huang et al. (1994)). (Open symbols based on breakthrough profile, closed symbols based on desorption profile.) Lines follow the closed symbols and are to guide the eye. c-d) IAS-theory prediction for the loading of (c) propane and (d) propylene for a (50:50) mixture on NaX.

113


paraffins when the partial pressure is increased or the temperature is decreased, except for propane, where the adsorbed amount shows an increase when the temperature is increased to 408 K at the higher pressures, as was also observed in the experimental data.

Based on the calculated adsorbed amounts the selectivities of the adsorbents were calculated for the binary mixtures. In Figs. 5.25a-b and 5.26a-b the experimental mixture selectivity of the two adsorbents for both the ethylene/ethane mixtures and the propylene/propane mixtures is plotted over the experimental pressure range. Because of the error in the calculated adsorbed amounts of the paraffins based on the breakthrough profiles, as expected a large difference with the mixture selectivity calculated from the desorption profiles is seen. As indicated in Fig. 5.27 a large relative error is present at the lower pressures.

The selectivity of ethylene on 36 wt% CuCl/NaX shows a decreasing trend when the partial pressure is increased (Fig. 5.25b), except at 318 K were an increase is observed at higher pressures. At higher temperatures the selectivity of the 36 wt% CuCl/NaX adsorbent for ethylene increases. At the lower pressure regime a selectivity of about 50-200 is found, while at the higher pressures the selectivity is approximately 15-20 for ethylene. For the large

Pressure [kPa]

1

Sel

ectiv

ity [-

]

102

102

10110-1

10

a NaX

318 K

358 K408 K

Pressure [kPa]

1

Sel

ectiv

ity [-

]

102

102

10110-1

10

a NaX

318 K

358 K408 K

1

Pressure [kPa]

Sel

ectiv

ity [-

]

102

103

10110-1

10

102

b

36 wt% CuCl/NaX

318 K

358 K

408 K

1

Pressure [kPa]

Sel

ectiv

ity [-

]

102

103

10110-1

10

102

b

36 wt% CuCl/NaX

318 K

358 K

408 K

1

Pressure [kPa]

Sel

ectiv

ity [-

]

102

102

10110-1

10318 K

358 K408 K

c NaX

1

Pressure [kPa]

Sel

ectiv

ity [-

]

102

102

10110-1

10318 K

358 K408 K

1

Pressure [kPa]

Sel

ectiv

ity [-

]

102

102

10110-1

10318 K

358 K408 K

c NaX

1

Pressure [kPa]

Sel

ectiv

ity [-

]

102

103

10110-1

10

102

318 K

358 K

408 K

d

36 wt% CuCl/NaX1

Pressure [kPa]

Sel

ectiv

ity [-

]

102

103

10110-1

10

102

318 K

358 K

408 K

1

Pressure [kPa]

Sel

ectiv

ity [-

]

102

103

10110-1

10

102

318 K

358 K

408 K

d

36 wt% CuCl/NaX

Fig. 5.25: a-b) Selectivity for ethylene over (a) NaX and (b) 36 wt% CuCl/NaX at 318 K ( , ), 358 K ( ,and 408 K ( , ).(Open symbols based on breakthrough profile, closed symbols based on desorption profile.) Lines follow the closed symbols and are to guide the eye. c-d) IAS-theory estimation for the selectivity forethylene over (c) NaX and (d) 36 wt% CuCl/NaX for a (50:50) mixture with ethane.

114

Chapter 5

36 wt% CuCl/NaX column a selectivity for ethylene of 25 was obtained at 408 K and a partial pressure of 27 kPa for both components, which is in close agreement with the values obtained for the smaller 36 wt% CuCl/NaX-column.

On the NaX adsorbent (Fig. 5.25a) the selectivity of ethylene is lower than on the 36 wt% CuCl/NaX adsorbent. At 318 K the selectivity for ethylene decreases from 10 to 4 when the partial pressure is increased, while at the other temperatures the selectivity remains almost constant at 3-4.

The estimations of the mixture selectivity, based on the IAS-theory, are presented in Fig. 5.25c-d for ethylene/ethane mixtures. On both adsorbents a decrease in the selectivity is predicted when the temperature is increased. For the NaX adsorbent the selectivity remains fairly constant over the experimental pressure range, in agreement with the experimental data. At 318 K a small decrease is observed at higher pressures. For the 36 wt% CuCl/NaX adsorbent the IAS predictions of the mixture selectivity are more than a factor 10 higher than for the NaX adsorbent, as is approximately in agreement with the experimental data. Similarly as the experimental data, an increase in the partial pressure results in a decrease in the

1

Pressure [kPa]

Sel

ectiv

ity [-

]

102

102

10110-1

10

a NaX

318 K

358 K

408 K

1

Pressure [kPa]

Sel

ectiv

ity [-

]

102

102

10110-1

10

a NaX

318 K

358 K

408 K

1

Pressure [kPa]

Sel

ectiv

ity [-

]

102

102

10110-1

10

b

36 wt% CuCl/NaX

318 K358 K408 K

1

Pressure [kPa]

Sel

ectiv

ity [-

]

102

102

10110-1

10

b

36 wt% CuCl/NaX

318 K358 K408 K

1

Pressure [kPa]

Sel

ectiv

ity [-

]

102

102

10110-1

10318 K358 K408 K

c NaX

1

Pressure [kPa]

Sel

ectiv

ity [-

]

102

102

10110-1

10318 K358 K408 K

1

Pressure [kPa]

Sel

ectiv

ity [-

]

102

102

10110-1

10318 K358 K408 K

c NaX

1

Pressure [kPa]

Sel

ectiv

ity [-

]

102

102

10110-1

10

318 K

358 K

408 K

d

36 wt% CuCl/NaX1

Pressure [kPa]

Sel

ectiv

ity [-

]

102

102

10110-1

10

318 K

358 K

408 K

1

Pressure [kPa]

Sel

ectiv

ity [-

]

102

102

10110-1

10

318 K

358 K

408 K

d

36 wt% CuCl/NaX

Fig. 5.26: a-b) Selectivity for propylene over (a) NaX and (b) 36 wt% CuCl/NaX at 318 K ( , ),358 K ( , and 408 K ( , ). ( ) Literature data for propylene on NaX at 343 K (Huang et al. (1994)). (Open symbols based on breakthrough profile, closed symbols based on desorption profile.) Lines follow the closed symbols and are to guide the eye. c-d) IAS-theory estimation for the selectivity for propylene over (c) NaX and (d) 36 wt% CuCl/NaX for a (50:50) mixture with propane.

115


selectivity of the 36 wt% CuCl/NaX adsorbent in the lower pressure range, while at the higher pressures and lower temperatures an increase in the selectivity is observed.

The propylene selectivity on 36 wt% CuCl/NaX remains relatively constant up to a partial pressure of 27 kPa (Fig. 5.26b). Thereafter a decrease to a mixture selectivity of 7-8 was found in the experimental data. At the lowest temperature (318 K) the highest mixture selectivity for propylene (25-30) was obtained for this adsorbent. At the other temperatures the mixture selectivity for propylene is approximately 15-20.

On the NaX adsorbent the propylene mixture selectivity is lower compared to the 36 wt% CuCl/NaX adsorbent. A small increase at 318 K and a decrease at 408 K are observed when the partial pressure is increased within the measurement interval (Fig. 5.26a). Within the experimental pressure and temperature range the mixture selectivity for propylene varies between 2½ and 7.

Application of the IAS-theory to the propane and propylene isotherms results in the mixture selectivities presented in Fig. 5.26c-d. In agreement with the experimental data, on the NaX-adsorbent the selectivity remains almost constant over the experimental pressure range. Only a small increase is observed, especially at the lower temperatures. On the 36 wt% CuCl/NaX adsorbent the selectivity for propylene is approximately a factor 2 higher. Increasing the total pressure of a (50:50) mixture or decreasing the temperature results in a (small) increase in the mixture selectivity.

10-4

Pressure [kPa]

Rel

ativ

e er

ror [

-]

102

10

10110-1

10-3

10-2

10-1

1

36 wt% CuCl/NaX

C3H8 & selectivity

C3H8 & selectivity

C3H6

C3H6

10-4

Pressure [kPa]

Rel

ativ

e er

ror [

-]

102

10

10110-1

10-3

10-2

10-1

1

36 wt% CuCl/NaX

C3H8 & selectivity

C3H8 & selectivity

C3H6

C3H6

Fig. 5.27: Relative standard deviation of the loading of propane ( , ) and propylene ( , ) and relative standard deviation of the selectivity ( , ) for the breakthrough profile ( , , and dotted line) and the desorption profile ( , , and solid line) on 36 wt% CuCl/NaX at 358 K. Lines are indicative and to guide the eye.

5.5 Discussion The breakthrough and desorption profiles in Fig. 5.5a-b show a good reproducibility and

therefore it is concluded that the performance of the adsorbents remains stable during the experiments. The fast analysis of the Compact GC allows the capture of the complete breakthrough and desorption profiles for the binary mixtures with only two duplicated experiments and the analysis shows a good reproducibility.

116

Chapter 5

As expected, a two times longer breakthrough time is obtained once the hydrocarbon feed flowrate is 50% smaller (Fig. 5.6a) compared to the later experiments (runs K and L in Table 5.4 at 358 K). It shows that the equilibrium adsorbed amount is not affected by the flowrate of the gas mixture, as long as the pressure drop is small.

The similar breakthrough times of the 36 wt% CuCl/NaX column in Fig. 5.5b compared to the other experiments (runs M and N for the small 36 wt% CuCl/NaX column) shows that also this adsorbent can be prepared with good reproducibility. The presence of 100 kPa of helium (on a total pressure of 200 kPa) in this experiment apparently did not significantly affect the obtained breakthrough profiles and equal adsorbed amounts are obtained, since the partial pressures of the two components were the same for both experiments.

5.5.1 36 wt% CuCl/NaX adsorbent In Figs. 5.6 and 5.11a-b a very early breakthrough is seen for ethane on 36 wt%

CuCl/NaX. This indicates that only a minor amount of ethane is adsorbed by the adsorbent. This is confirmed in Fig. 5.10 where the integrated amount of ethane is relatively small. The later breakthrough of ethylene confirms the stronger affinity of the adsorbent for the olefin. At lower temperatures the ethylene breakthrough occurs earlier than at higher temperature, which is in contradiction with the trend observed in the single component isotherms reported earlier (Chapter 4 of this thesis), which show higher adsorption capacities at lower temperatures. On the other hand a long time is required to reach the final equilibrium composition in the later segments of the breakthrough profiles, which suggests that the complete saturation of the adsorbent with ethylene occurs much later. This slow approach to the feed composition can be explained by the presence of diffusion limitations. Within a given time interval, the slow diffusion process reduces the adsorption of the olefins. Therefore the mol fraction in the gas phase will remain relatively high and breakthrough of the olefin will occur relatively early. Thereafter the mol fraction slowly increases, while more gas counterdiffuses into and out of the pores of the zeolite until the binary equilibrium between the adsorbent and gas phase is obtained. Indicative of the presence of diffusion limitations was also that at the higher temperatures a much shorter equilibration time was required.

At a low partial pressure and high flowrates the breakthrough profiles broaden as seen in Figs. 5.12 and 5.14b. As expected, the higher flowrate results in a shorter time delay, because of the shorter residence time in the column. However, the observed broadening might be surprising, but can be explained as follows. The higher velocity of the gas through the column (254 ml min-1 (SATP)) results in a significant increase of the pressure drop across the column to ~ 30 kPa. Once the flow through the column is switched from helium (20 ml min-1 (SATP)) to the gas mixture (254 ml min-1 (SATP)), the pressure in the entrance part of the column will increase because of this higher pressure drop and since the outlet pressure is fixed. Part of the gas mixture is ‘stored’ in this part of the column during this pressure increase. Furthermore, since the gas density will increase and the molar flow is fixed, the volumetric flowrate will be smaller than 254 ml min-1. This phenomenon results in the broader breakthrough profiles shown in Fig. 5.14b, instead of the expected sharper breakthrough profile at higher velocities and Péclet numbers.

117


Because of the larger pressure drop across the column, the equilibrium amount adsorbed at the top of the column will be larger than that at the bottom of the column. When the flow is reduced from 254 to 20 ml min-1 (SATP) for the desorption experiment, the pressure drop across the column will be smaller. This will result in a relatively fast decrease of the partial pressures in the top of the column. The adsorbent will then no longer be in equilibrium with the gas phase and desorption will start immediately. As a consequence, the mol fraction of primarily the olefin will initially increase beyond the original feed composition. This phenomenon explains the initial increase of the mol fraction seen in Fig. 5.13. As expected, the desorption profiles for SiC (Fig. 5.14d), do not show this increase, since the adsorbed amounts will be zero on this ‘inert’ material.

Because of this pressure drop at the lower partial pressures the obtained values for the absorbed amount will be slightly higher than expected based on the backpressure of the column and feed composition. The average partial pressure of both components can be approximately 15% higher for a mol fraction of 0.00788. At the higher concentrations the pressure drop is much smaller (~ 1 kPa for 8 ml min-1 (SATP)) and the error is the partial pressure for a mol fraction of 0.25 will only be 1%. The absolute error in the partial pressures will therefore be relatively small for all concentrations used in this study. Its effect on the adsorbed amounts will therefore be relatively small.

In Fig. 5.15a-c the breakthrough of propane occurs later than for ethane (Figs. 5.6 and 11a-b), which suggests that more propane is initially adsorbed on the 36 wt% CuCl/NaX adsorbent. However during the breakthrough experiment an interesting phenomenon occurs along the column. In the beginning of the breakthrough experiment all propylene is adsorbed in the top part of the column. Therefore, the partial pressure of propane will initially be higher in the bottom part of the column. This higher partial pressure, and the absence of propylene, will result in an initially higher adsorption of propane in the adsorbent even beyond the binary equilibrium value that can be expected based on the feed composition. As a consequence the breakthrough of propane will be further postponed, as seen in the breakthrough profiles. During the breakthrough experiment the propylene front will advance through the column and binary equilibrium will be established. Therefore (part of) the earlier adsorbed propane will have to desorb from the adsorbent. This desorption results in a roll-up of the mol fraction of propane beyond the value of 1/3 expected based on the feed composition (see Eq. 5.8), as is indeed observed in the breakthrough profiles. Because of this desorption, the final amount of propane adsorbed on the adsorbent will be lower than expected based on its breakthrough time. This will finally result in a higher mixture selectivity of the adsorbent for propylene.

The breakthrough time and the amount adsorbed is only 2½ times larger for the larger 36 wt% CuCl/NaX column than for the small column, while the total amount of adsorbent is 5 times larger. This would suggest different properties of the adsorbent in this column, since the two smaller 36 wt% CuCl/NaX columns used in this study gave comparable results. Indeed, the packings are different. For the larger column larger zeolite crystals of ~100-120 m were used. The diffusion length in these particles is longer and a longer equilibration time could be required. Also, for the larger zeolite crystals any blockage of pores in the outer part of the

118

Chapter 5

particle will make a large volume of the zeolite crystals inaccessible. These blockages could occur due to imperfections in the zeolite structure, the presence of CuCl particles or the presence of adsorbed paraffin in the interior of the crystal. Finally variations in the outcome of the synthesis could have resulted in a different or uneven distribution of the CuCl inside the pores of the larger zeolite crystal compared to the smaller crystals. For instance, the larger crystal may need a longer heat treatment time for the diffusion and dispersion of CuCl inside the zeolite pores. Despite the smaller amount adsorbed on the larger column, the profiles of the breakthrough and desorption profiles do show a large similarity with those of the smaller column with smaller crystals. Furthermore, the mixture selectivity of the accessible part is 25 at 408 K and 27 kPa, which is in close agreement with the results obtained with the smaller column, so the difference is attributed to a lower effective capacity of the large crystals.

5.5.2 NaX adsorbent The paraffins show a later breakthrough on the NaX adsorbent than on the 36 wt%

CuCl/NaX adsorbent, which indicates that a significant amount of the paraffin is initially adsorbed on the NaX-adsorbent. However, the roll-up of the mol fraction of the paraffin far above 1/3 suggests that a considerable amount is desorbed, while the olefin progresses through the column. This displacement of the paraffins will finally result in a lower amount of the adsorbed paraffins and therefore a higher mixture selectivity is achieved than expected based on the breakthrough times only. The later breakthrough time of the olefins compared to the 36 wt% CuCl/NaX adsorbent indicates that the adsorbed amount of the olefins is larger for the NaX adsorbent. The main reason for this difference is the presence of CuCl in the pores of the zeolite, which reduces the volume that is available for the adsorption of the olefin.

The presence of diffusion limitations is considerably reduced on the NaX column, and as a consequence, the breakthrough profiles of the olefins are much steeper and the equilibrium appears to be established relatively fast. At the higher temperatures the breakthrough times are shorter, as can be expected based on the single component adsorption behaviour, which shows a lower adsorption capacity for higher temperatures (Chapter 4 of this thesis).

The interruption in the steep decrease of the mol fraction of the olefins observed in many of the desorption profiles with NaX at high partial pressures (seen for example after ~ 250 seconds in Fig. 5.18) can be explained by re-adsorption of the olefins. For the NaX adsorbent a relatively large amount of the paraffin is adsorbed at binary equilibrium. Because of the relatively fast diffusion in the NaX zeolite, the paraffins can quickly desorb from the adsorbent, after which the empty adsorption site can be re-occupied by olefins that were desorbed in an earlier section of the column. The fraction of olefin leaving the column will then initially be lower than without this effect, until all paraffin has been removed from the column. At that moment the adsorbed amount in the bottom part of the column shows a short interruption and will then continue to decrease in a similar trend as without this effect.

119


5.5.3 Single component breakthrough The selectivity of the adsorbents can also be seen in the breakthrough profile of the single

components. Although their purity in the gas cylinder is high, traces of paraffins (or olefins) are present. Due to the selective adsorption of the olefin, the small paraffin impurity in the olefin gas cylinder is concentrated along the column. This paraffin impurity will then break through first. After the breakthrough of the olefin, the outlet composition will return to the supply specification of the olefin gas cylinder. For NaX the increase in the paraffin mol fraction is more pronounced than on the 36 wt% CuCl/NaX adsorbent, because of the replacement of the paraffin by the olefin on the NaX adsorbent (roll-up effect) and the larger adsorption capacity of the NaX adsorbent.

The olefin impurities in the paraffin gas cylinders are completely adsorbed on the 36 wt% CuCl/NaX adsorbent. The higher selectivity of this adsorbent for the olefins makes it an ideal adsorbent to remove an olefin impurity from a paraffin feed. Because of the lower selectivity of the NaX adsorbent, the breakthrough of ethylene occurs at 358 and 408 K. Only at the lower temperature (318 K) the maximum adsorbed amount of ethylene is not surpassed during the time of the experiment. Assuming all the olefins fed to the column are adsorbed in the column, it is expected to breakthrough after ~ 2 h.

For propane the breakthrough of the propylene impurity is not seen for both adsorbents at the experimental temperatures. As was seen earlier, propylene has a larger affinity with NaX than ethylene and therefore more propylene can be adsorbed on this adsorbent and prevents its breakthrough. Furthermore the purity of the propane gas cylinder is higher (99.95% instead of 99.9% for ethane), so less propylene is fed to the column. The adsorbed amount is large enough to prevent its breakthrough during the experimental time of 1 h.

5.5.4 Adsorption capacities and selectivities The trends of the adsorbed amounts presented in Figs. 5.21-5.24 showed a large similarity

with those predicted with the IAS-theory. In agreement with this theory both samples show a relatively high selectivity for the olefins. For the NaX sample also the quantitative values of the integral adsorbed amounts and the selectivity corresponded well with the predictions of the IAS-theory. For the 36 wt% CuCl/NaX sample larger differences were observed, especially for propane and propylene, though most of the trends are similar. At a low temperature the adsorbed amounts and so the selectivity are influenced by diffusion limitations for this adsorbent, while at a higher temperature the small adsorbed amounts of especially the paraffins resulted in larger experimental errors. Furthermore, the adsorption of the olefins via -complexation with CuCl could cause a less ideal adsorbed phase than is assumed in the IAS-theory.

The negative values of the adsorbed amounts obtained for some of the experiments could be the effect of a small difference in the porosity of the adsorption columns and the SiC-column. In case the porosities of the SiC-column and the other columns are the same, the corrected results will still be close to the real adsorbed amounts. However, a small difference in the porosity will result in larger errors in the adsorbed amounts at the higher partial

120

Chapter 5

pressures, since the neglected sum in the nominator of Eq. 5.3 becomes more important. At the lower pressure the adsorbed amounts will be relatively small, so a small absolute error could already result in a negative value, when correcting with the outcome of the SiC-column, especially for the values obtained from the breakthrough profiles. As shown in Fig. 5.27 the relative error of the adsorbed amounts of the paraffins obtained from the breakthrough profiles becomes large at the lower pressures.

For most pressures and temperatures the dispersion of CuCl in the pores and cavities of the NaX crystals reduces the adsorbed amount of both components. Only at the higher temperatures the adsorbed amount of the olefins on the 36wt% CuCl/NaX sample is comparable with the zeolite NaX sample. Part of the internal volume of the zeolite will be occupied by CuCl, which reduces the available pore volume for adsorption. For the paraffin, which can only adsorb via the weak physical Van der Waals forces, the CuCl dispersion therefore reduces the adsorption capacity, while for the olefins a lower decrease is observed, which is attributed to the stronger interaction of the olefins with CuCl via a -complex. This effect (partly) compensates the decrease in the adsorption capacity, caused by the smaller pore volume available for adsorption. Since this -complex remains relatively strong at the higher temperatures, the adsorbed amounts of the olefins will remain higher at these conditions.

The dispersion of CuCl on the walls of the pore openings and cavities in the NaX zeolite decreases their size and the available free space in these cavities. So not only the capacity is reduced, but also the transport of the gases from the surface towards the centre of the crystals is strongly hampered. This was also observed earlier for the measurements of the adsorption isotherms (Chapter 4 of this thesis). Attainment of adsorption equilibrium for the CuCl/NaX sample took much longer than for NaX. The presence of diffusion limitations has a considerable effect on the achieved loadings on the 36 wt% CuCl/NaX adsorbent and the breakthrough profiles at the lower temperatures. This activated diffusional transport in the pores of the zeolite is increased whenever the temperature is increased, alleviating diffusion limitations. Indeed shorter equilibration times were required at higher temperatures.

The smaller size of the pore openings and cavities will be more important if a larger molecule is adsorbing. For propane and propylene the adsorbed amounts are therefore more affected by diffusion limitation than for ethane and ethylene. The attainment of the adsorption equilibrium in the breakthrough profiles did indeed require a much longer equilibration time for propane and propylene mixtures.

The adsorbed amounts of both components could also be affected by the presence of the other components in the mixture. During a breakthrough experiment initially (most of) the olefins will be adsorbed in the top part of the breakthrough column. Therefore, initially only the paraffins (in helium) will be present in the bottom part of the column. These paraffins will partly diffuse into the pores of the adsorbents and adsorb. During the breakthrough experiment the olefins will slowly advance through the column. To achieve binary equilibrium part of the paraffins will desorb and will have to diffuse out of the zeolite countercurrently with the entering olefin. Part of the paraffins could then be locked in the centre or a pocket of the zeolite crystals and will be unable to desorb from the zeolite crystal.

121


These locked paraffins would result in a lower mixture selectivity of the adsorbent than expected for the true binary equilibrium (Mittelmeijer-Hazeleger, et al. (2002)).

On the NaX adsorbent the adsorbed amounts of ethylene and ethane seem not to be affected by diffusion limitations. A lower temperature or a higher partial pressure indeed results in higher adsorption capacities (Figs. 5.23a-b). The higher temperature reduces the adsorption affinity with the adsorbent and therefore a smaller amount will be adsorbed.

For propylene and propane on NaX an increase in the temperature results in a decrease in the adsorbed amount of propane at the lower pressures, whereas it increases at the higher pressures (Fig. 5.24a). A similar transition is observed for ethane on 36 wt% CuCl/NaX (Fig. 5.21a). These observations are in agreement with the predictions of the IAS-theory. At the lower pressures the adsorption is mainly determined by the adsorption affinity. Higher temperatures will reduce the adsorptive-adsorbent interactions and therefore a decrease in the adsorbed amount is observed at these lower pressures. At the higher pressures the adsorption becomes also affected by entropic effects. The component that has the highest packing efficiency on the adsorbent will be forced to occupy all the adsorption sites, expelling the other component (Kapteijn, Moulijn, and Krishna (2000); Krishna and Baur (2003); Krishna and Paschek (2000)). At the lower temperatures the loadings increase and the importance of the packing efficiency becomes more important and the transition from enthalpy to entropy control will occur at lower pressures (Fox and Bates (2004); Zhu et al. (2005)). Therefore the loading of propane on NaX still increases with pressure at 408 K, while it becomes constant at the lower temperatures. Similarly the loading of ethane on 36 wt% CuCl/NaX decreases at higher pressures at 318 K. This is one of the first experimental data were this transition in the adsorbed amounts is indeed observed.

On Faujasite (NaY) an entropy/packing effect was observed earlier in the literature for other olefin/paraffin mixtures (Daems et al. (2005); Denayer, Daems, and Baron (2006)). The liquid-phase adsorption of olefin/paraffin mixtures of different lengths (e.g. hexene/decane and dodecene/decane mixtures) showed a larger selectivity for the shorter olefin, compared to the longer olefin, at these high loadings, since the shorter olefin is able to pack more efficiently next to the long paraffin in the super cavity of Faujasite.

The mixture selectivities presented in Figs. 5.25a-d and 5.26a-d indicate that both adsorbents show a selective adsorption of the olefins. Compared to the NaX zeolite, the dispersion of CuCl in the pores and cavities increases the selectivity for the olefins. Depending on the temperature and partial pressures, the mixture selectivity for ethylene is increased from 3-10 to 15-200 and for propylene from 2.5-7 to 15-30. Especially at the lower pressures the increase in the selectivity is considerable, once CuCl is present in the pores. This observation is in agreement with the strong -complex of CuCl and the olefin and with our earlier reported isotherm data (Chapter 4 of this thesis). The -complex is already formed at relatively low partial pressures. The adsorption constant of the olefin on 36 wt% CuCl/NaX adsorbent is significantly larger than that of the paraffin. As a consequence, the adsorbed amount of the olefins is already very high at the lower pressures and a large selectivity is expected. For the binary system this preference results in a larger adsorbed amount for the

122

Chapter 5

olefins at lower pressures and therefore a higher selectivity at these lower pressures is achieved.

The selectivity of the NaX adsorbent remains very constant over the investigated pressure range, which shows that the operational pressure is not important for the selectivity of NaX within the current experimental pressure range. Although the predicted selectivities are approximately 2-3 times larger than the experimental data, the increasing and decreasing trends with partial pressure are in good agreement with the IAS-theory. Furthermore the trend and the value of the propylene selectivity is in close agreement with the literature values of 11 (at 298 K), 9 (at 323 K) and 8 (at 343 K) for a binary (50:50) propylene/propane mixture (Huang et al. (1994)). Due to the effects of the entropy and the packing efficiency, the mixture selectivity of propylene on NaX and ethylene on 36 wt% CuCl/NaX at 318 K increases with pressure, as is in agreement with the IAS-theory.

The selectivity of the 36 wt% CuCl/NaX adsorbents shows a larger deviation from the predictions of the IAS-theory, especially for propylene/propane mixtures. This may be attributed to diffusion limitations and non-ideal adsorption behaviour of the -complex. Despite the deviations the values of the predicted selectivities have a similar order of magnitude as those predicted by the IAS-theory.

For the application of the adsorbents for olefin/paraffin separation, a choice between the higher selectivity of the 36wt% CuCl/NaX adsorbent and the higher adsorption capacity and faster diffusion of the NaX adsorbent has to be made. It should be taken into account that the diffusion limitation of the 36 wt% CuCl/NaX adsorbent can be reduced by the use of smaller particles, though for zeolite membrane operation a thinner membrane may affect the membrane quality. The separation of the smaller molecules, ethylene and ethane, benefits more from the dispersion of CuCl and very high selectivities are obtained. In industrial practice the separation of propylene/propane in a pressure swing adsorption (PSA) unit would probably be more efficient with a NaX adsorbent without dispersed CuCl. For an ethylene/ethane mixture the separation can be performed in a PSA operation with the CuCl/NaX adsorbent. However, the stronger affinity would require the operation at higher temperatures than for NaX, otherwise desorption times can be very long. Alternatively, a vacuum swing PSA operation or a temperature swing adsorption (TSA) can be considered for the CuCl/NaX adsorbent to separate these mixtures.

5.6 Conclusions The fast analysis of the microGC allowed the recording of many breakthrough and

desorption profiles in a short time and with excellent reproducibility. The NaX and 36 wt% CuCl/NaX samples showed a selective adsorption of the olefins for an ethylene/ethane and a propylene/propane mixture. During the loading of the breakthrough column a displacement of the adsorbed paraffins by the olefins occurred along the length of the column. This roll-up effect was more pronounced on the NaX sample, since a significant amount of the paraffins was initially adsorbed during the loading of the column with a binary olefin/paraffin mixture.

123


On the 36 wt% CuCl/NaX column only a small amount of the paraffins was adsorbed and therefore this roll-up effect was very minor.

For the 36 wt% CuCl/NaX adsorbent, the dispersion of CuCl inside the zeolite crystal results in smaller pore sizes and pore volume, lowering the adsorption capacity and causing diffusion limitations. The latter affected the profiles and resulted in errors in the calculation of the adsorbed amounts at the binary equilibrium, especially for propane and propylene. At higher temperatures the diffusion limitations were reduced. On the NaX sample, diffusion limitations were not observed and sharper breakthrough profiles were obtained than on 36 wt% CuCl/NaX.

The binary isotherm data of NaX corresponded well with the predictions of the IAS-theory using the single component isotherms recorded earlier, both qualitatively and quantitatively. The selectivity on NaX remained constant over the entire experimental pressure range. For the 36 wt% CuCl/NaX adsorbent larger differences with the predictions of the IAS-theory are found, though the qualitative trends in the adsorbed amounts corresponded well. Deviations could be the effect of the diffusion limitations or the non-ideal adsorption via the -complex with CuCl.

The binary isotherms of propylene and propane on NaX showed that at the lower partial pressures the adsorption is primarily controlled by the adsorption affinity of the adsorbent with propylene. At the higher partial pressure the adsorption is primarily controlled by the entropy and the molecule with the most effective packing efficiency (propylene) is adsorbed preferentially. In agreement with the IAS-theory a similar transition was observed for ethylene/ethane mixtures on 36 wt% CuCl/NaX.

The dispersion of CuCl in the pores of the zeolite resulted in an increase of the selectivity from 3-10 for ethylene on NaX to 15-200 on 36 wt% CuCl/NaX, and from 3-7 for propylene on NaX to 15-30 on 36 wt% CuCl/NaX.

Because of the diffusion limitations and the strong affinity of the 36 wt% CuCl/NaX adsorbent, the application of the adsorbent in a PSA separation process would require the operation at either higher temperatures or lower pressures. Alternatively TSA operation may be considered. Since the adsorption capacities are also considerably reduced at these conditions, the NaX adsorbent may be preferred for these separations.

5.7 List of symbols

5.7.1 Variables

mol, in Molar flowrate into the column [mol s-1]v, in Volumemetric flowrate into the column [m3 s-1]v,helium,in Volumemetric flowrate of helium into the column [m3 s-1]v, out Volumemetric flowrate out of the column [m3 s-1]v,ethane,out Volumemetric flowrate of ethane out of the column [m3 s-1]

Ptot Total pressure at the bottom of the column [Pa]qi Adsorbed amount of component i [mol kg-1]Rg Universal gas constant [J mol-1 K-1]

124

Chapter 5

T Column temperature [K] t Time [s]tstart Start time of the breakthrough or desorption experiment [s] V Volume of the column [m3]Vmol Molar volume of the gas [m3 kmol-1]xi,in Mol fraction of component i in the feed [-]xi,out Mol fraction of component i in the outlet [-]

ix Average mol fraction in the column [-]

Porosity of the adsorbent bed in the column [-] Density of the adsorbent [kg m-3]

5.7.2 Abbreviations SATP Standard Ambient Temperature and Pressure (298 K and 101 kPa) STP Standard Temperature and Pressure (273.15 K and 101.325 kPa)

5.7.3 Flow sheet abbreviations BPC1 Back pressure controller 1 Cond1 Condense trap 1 CV1 Check valve 1 FIA1 Flow indicator 1 to trigger alarm if SRV’s open FI2 Flow indicator 2 FIDs Flame ionisation detectors Filt1 Filter 1MAN1 Manifold 1MFC1 Mass flow controller 1 PC1 Manual pressure controller 1 SRV1 Safety relieve valve 1 SV1 Solenoid valve 1 TC Temperature sensor for controller TS Temperature sensor for safety V1 Valve 1

P Differential pressure sensor

5.8 References Al-Baghli, N. A. and Loughlin, K. F., Binary and Ternary Adsorption of Methane, Ethane,

and Ethylene on Titanosilicate ETS-10 Zeolite, J. Chem. Eng. Data 51 (2006) 248-254. Bárcia, P. S., Silva, J. A. C. and Rodrigues, A. E., Separation by Fixed-Bed Adsorption of

Hexane Isomers in Zeolite BETA Pellets, Ind. Eng. Chem. Res. 45 (2006) 4316-4328. Brandani, F. and Ruthven, D. M., Measurement of Adsorption Equilibria by the Zero Length

Column (ZLC) Technique Part 2: Binary Systems, Ind. Eng. Chem. Res. 42 (2003) 1462-1469.

125


Daems, I., Leflaive, P., Méthivier, A., Denayer, J. F. M. and Baron, G. V., A Study of Packing Induced Selectivity Effects in Liquid Phase Adsorption of Alkane/Alkene Mixtures on NaY, Micropor. Mesopor. Mat. 82 (2005) 191-199.

Denayer, J. F. M. and Baron, G. V., Adsorption of Normal and Branched Paraffins in Faujasite Zeolites NaY, HY, Pt/NaY and USY, Adsorption 3 (1997) 251-265.

Denayer, J. F. M., Daems, I. and Baron, G. V., Adsorption and Reaction in Confined Spaces, Oil Gas Sci. Technol. 61 (2006) 561-569.


Fox, J. P. and Bates, S. P., Simulating the Adsorption of Binary and Ternary Mixtures of Linear, Branched, and Cyclic Alkanes in Zeolites, J. Phys. Chem. B 108 (2004) 17136-17142.

Ghosh, T. K., Lin, H.-D. and Hines, A. L., Hybrid Adsorption-Distillation Process for Separating Propane and Propylene, Ind. Eng. Chem. Res. 32 (1993) 2390-2399.

Grande, C. A. and Rodrigues, A. E., Propane/Propylene Separation by Pressure Swing Adsorption using Zeolite 4A, Ind. Eng. Chem. Res. 44 (2005) 8815-8829.

Hagemeyer, A., Jandeleit, B., Lie, Y., Poojary, D. M., Turner, H. W., Volpe Jr., A. F. and Weinberg, W. H., Applications of Combinatorial Methods in Catalysis, Appl. Catal. A-Gen. 221 (2001) 23-43.

Harlick, P. J. E. and Tezel, F. H., A Novel Solution Method for Interpreting Binary Adsorption Isotherms from Concentration Pulse Chromatography Data, Adsorption 6 (2000) 293-309.

Herberhold, M., Metal Pi-Complexes: Part II: Specific Aspects, Elsevier, New York (1974). Huang, Y.-H., Johnson, J. W., Liapis, A. I. and Crosser, O. K., Experimental Determination

of the Binary Equilibrium Adsorption and Desorption of Propane-Propylene Mixtures on 13X Molecular Sieves by a Differential Sorption Bed System and Investigation of their Equilibrium Expressions, Separ. Technol. 4 (1994) 156-166.


Kapteijn, F., Moulijn, J. A. and Krishna, R., The Generalized Maxwell-Stefan Model for Diffusion in Zeolites:Sorbate Molecules with Different Saturation Loadings, Chem. Eng. Sci. 55 (2000) 2923-2930.

Keller, J. U., Iossifova, N. and Zimmermann, W., Volumetric-Densimetric Measurements of the Adsorption Equilibria of Binary Gas Mixtures, Adsorpt. Sci. Technol. 23 (2005) 685-702.

Keller, J. U. and Staudt, R., Gas Adsorption Equilibria, Experimental Methods and Adsorption Isotherms, Springer, New York (2005).

Krishna, R. and Baur, R., Modelling Issues in Zeolite Based Separation Processes, Sep. Purif. Technol. 33 (2003) 213-254.

Krishna, R. and Paschek, D., Separation of Hydrocarbon Mixtures using Zeolite Membranes: a Modelling Approach Combining Molecular Simulations with the Maxwell-Stefan Theory, Sep. Pur. Technol. 21 (2000) 111-136.

126

Chapter 5

Lee, J.-W., Park, J.-H., Han, S.-S., Kim, J.-N., Cho, S.-H. and Lee, Y.-T., Adsorption Equilibrium and Dynamics of C4 Olefin/Paraffin on Pi-Complexing Adsorbent, Sep. Sci. Technol. 39 (2004) 1365-1384.

Linders, M. J. G., van den Broeke, L. J. P., Kapteijn, F., Moulijn, J. A. and Van Bokhoven, J. J. G. M., Binary Adsorption Equilibrium of Organics and Water on Activated Carbon, AIChE J. 47 (2001) 1885-1892.

Malek, A. and Farooq, S., Effect of Velocity Variation on Equilibrium Calculations from Multicomponent Breakthrough Experiments, Chem. Eng. Sci. 52 (1997) 443-447.

Malek, A., Farooq, S., Rathor, M. N. and Hidajat, K., Effect of Velocity Variation due to Adsorption-Desorption on Equilibrium Data from Breakthrough Experiments, Chem. Eng. Sci. 50 (1995) 737-740.

Mittelmeijer-Hazeleger, M. C., Ferreira, A. F. P. and Bliek, A., Influence of Helium and Argon on the Adsorption of Alkanes in Zeolites, Langmuir 18 (2002) 9613-9616.


Peng, J., Ban, H., Zhang, X., Song, L. and Sun, Z., Binary Adsorption Equilibrium of Propylene and Ethylene on Silicalite-1: Prediction and Experiment, Chem. Phys. Lett. 401 (2005) 94-98.

Sakuth, M., Meyer, J. and Gmehling, J., Measurement and Prediction of Binary Adsorption Equilibria of Vapors on Dealuminated Y-Zeolites (DAY), Chem. Eng. Process 37 (1998) 267-277.

Senkan, S., Ozturk, S., Krantz, K. and Onal, I., Photoionization Detection (PID) as a High Throughput Screening Tool in Catalysis, Appl. Catal. A-Gen. 254 (2003) 97-106.

Smit, B. and Krishna, R., Molecular Simulations in Zeolite Process Design, Chem. Eng. Sci. 58 (2003) 557-568.


Zhu, W., Groen, J. C., Van Miltenburg, A., Kapteijn, F. and Moulijn, J. A., Comparison of Adsorption Behaviour of Light Alkanes and Alkenes on Kureha Activated Carbon, Carbon 43 (2005) 1416-1423.

127


128

6Light olefin/paraffin separation:

Summary, process options and evaluation

A summary of the results of the previous chapters of this thesis is given. Based on these results the possibilities and problems for industrial application of the adsorbents in several process options, including PSA, TSA and membrane operations, are discussed. Of the two adsorbents investigated in this thesis, the CuCl/Faujasite adsorbent showed a higher selectivity compared to the Faujasite (NaX) adsorbent, with a beneficial impact on the number of separation steps. Because of the stronger adsorption, the use of CuCl/Faujasite would require the operation at higher temperature or sub-atmospheric pressure. However at these conditions the adsorption loadings will be lower. Furthermore, the adsorption process is affected by diffusion limitations. These disadvantages complicate the application of CuCl/Faujasite in separation processes and make its application less advantageous. Therefore, the pure Faujasite adsorbent is considered to be a better adsorbent candidate and could for instance be used in PSA or membrane operations to separate mixtures of ethylene/ethane or propylene/propane.

Light olefin/paraffin separation: Summary, process options and evaluation

6.1 Summary Light olefins, like ethylene and propylene are important feedstocks for the chemical

industry. They are mainly obtained via thermal cracking processes combined with energy intensive cryogenic distillation of the produced olefin/paraffin mixtures. As less energy intensive alternative, this separation can be achieved via a selective adsorption process. Therefore it is of the utmost importance to find a cheap and effective adsorbent.

The central theme of this thesis is the development and optimization of an adsorbent for the selective separation of olefin/paraffin mixtures and to test the application of these adsorbents. To this purpose a zeolite based adsorbent (Faujasite NaX or NaY) was used. Large NaX zeolite crystals were synthesized (Chapter 2 of this thesis) to obtain a low pressure drop over the packed column in the application tests in a breakthrough setup. The use of large crystals will also reduce the number of modelling variables that determine the separation performance, since in compressed pellets of smaller crystals this will be affected by the additional intraparticle transport phenomena in the pellet. In these zeolite crystals CuCl was successfully dispersed. Saturation loadings of 36 wt% CuCl for NaX and 43 wt% for NaY were found, which corresponds to approximately 10 CuCl molecules per super cavity.

FTIR spectroscopy with CO as a probe molecule (Chapter 2 of this thesis) confirmed that a dispersion of CuCl in the zeolite occurred. Furthermore a strong adsorption of CO on CuCl via -complexation was seen. Adsorption of ethane, ethylene, propane and propylene in the low-pressure IR transmission cell confirmed that the adsorption on NaY zeolite is weak for all components. A pressure reduction can quickly remove the adsorbed gases from the Faujasite zeolite. On the CuCl/Faujasite adsorbent the olefins show a much stronger adsorption. Additional absorption bands were observed with IR spectroscopy, corresponding to the -complex of Cu+ with the double bond of the olefin. Complete desorption of the olefins from this complex occurred at much lower pressures or at higher temperatures than from the Faujasite zeolite itself.

Besides the potential of the adsorbent to separate the olefin/paraffin mixtures it is also important that the adsorbent remains stable for a long time. In industrial feeds traces of water vapour can be present in the gas mixture. It was shown that the exposure of the CuCl/Faujasite adsorbent to ambient air results in the gradual destruction of the adsorbent. CuCl was oxidized to CuCl2·3Cu(OH)2 (Atacamite) when exposed to both water vapour and oxygen (Chapter 3 of this thesis). An expansion of the Cu-material inside the zeolite pore structure occurs, due to the lower molar density (mol Cu per m3) of the oxidized Cu-material. The confinement within the zeolite pores resulted in the growth of fibres out of the outer pores of the zeolite crystal and/or in the formation of cracks in the zeolite crystal. A higher amount of CuCl in the pores and cavities of the zeolite resulted in a longer stability of the adsorbent. This can be explained from a reduced diffusion of water and oxygen into the zeolite due to the presence of CuCl. Above the saturation capacity some CuCl particles will still be present outside the zeolite crystal and can quickly react with water and oxygen. Therefore for maximum stability the loading of CuCl in the zeolite should be close to, but not exceeding the saturation capacity.

130

Chapter 6

The presence of water vapour, without the presence of oxygen, led to a decrease in the adsorption capacity of CO and other gases on the CuCl/Faujasite. Water covers the adsorption sites in the adsorbent and/or restricts the access of adsorptives to the zeolite pores. After the removal of the adsorbed water, by heat treatment, the fast adsorption of CO was restored. For industrial practice air exposure should therefore be minimized, while the presence of traces of water vapour only is not necessarily destructive for the CuCl/Faujasite adsorbent.

Single component adsorption isotherms revealed that both NaX and CuCl/NaX showed a higher affinity for the olefins (Chapter 4 of this thesis). The dispersion of CuCl in the Faujasite zeolite resulted in a decrease in the adsorption capacities of both the olefins and the paraffins. This can be explained by the reduction of the available pore volume when CuCl is dispersed on the walls of the zeolite pore structure. At lower pressures, higher (or less reduced) loadings of the olefins were found for the CuCl/Faujasite adsorbent compared to the Faujasite adsorbent, because of the higher affinity of the olefins for CuCl via a strong -complexation. About 17% of the Cu+ is involved in the -complexation. A larger adsorption constant and heat of adsorption was found for the olefins on CuCl/NaX. The measurement of the adsorption isotherms via the volumetric method also showed that the equilibration time was increased when CuCl is dispersed in the pores of the NaX, indicating a slower diffusion of the adsorptives in the adsorbent.

The selective adsorption of the olefins was confirmed in the breakthrough setup. Binary adsorption data were determined with this setup between 0.8-54 kPa at 318, 358 and 408 K (Chapter 5 of this thesis). On the CuCl/NaX adsorbent the adsorption of the hydrocarbons was affected by diffusions limitation. Below 358 K these diffusion limitations are more pronounced, especially for the relatively large propane and propylene molecules. At higher temperatures diffusion limitations are reduced, but also the adsorption capacities of the CuCl/NaX adsorbent are smaller. The highest selectivities were obtained at low (sub-atmospheric) partial pressures. For ethylene the dispersion of CuCl in the pores of the NaX zeolite results in a much larger increase of the selectivity compared to propylene. The mixture selectivity for ethylene increased from 3-10 on NaX to 15-200 on CuCl/NaX, while the mixture selectivity for propylene increased from 3-7 on NaX to 8-30 on CuCl/NaX. The smaller ethylene molecule diffuses much easier through the narrower pores towards the centre of the large zeolite crystal. Furthermore, the interaction of ethylene with CuCl via -complexation is much stronger than for propylene, since the propylene contains an extra methyl group. If the selectivity is independent of the crystal size, diffusion limitations can be reduced by the use of smaller crystals to allow a faster industrial process. To maintain a low pressure drop across the adsorption column, these smaller crystals can be pressed into larger particles.

On NaX, without CuCl, also a selective adsorption of the olefins was observed. The binary equilibrium was reached relatively fast after the breakthrough of both components, which suggest that the diffusion is much faster in the blank zeolite. The lower selectivity for NaX remains almost constant over the investigated pressure range (0.8-54 kPa). For a binary (50:50) mixture of propylene and propane a transition from an enthalpy controlled adsorption at lower loadings to an entropy affected adsorption at higher loadings was observed with

131


favourable affect on the selectivity. A similar transition and favourable effect on the selectivity was observed for ethylene/ethane (50:50) mixtures on 36 wt% CuCl/NaX. This is one of the first experimental data where this transition is indeed observed.

The mixture adsorption of the binary (50:50) olefin/paraffin mixture on NaX can be well described with the Ideal Adsorbed Solution (IAS) theory (Chapter 1 and 5 of this thesis) and the single component isotherms, both qualitatively and quantitatively. For the 36 wt% CuCl/NaX adsorbent larger differences with the predictions of the IAS-theory were found, probably because of the diffusion limitations or the non-ideal adsorption via -complexation; though the qualitative trends corresponded well. The observed transition from enthalpy controlled to entropy affected adsorption was indeed predicted with the IAS-theory.

6.2 Evaluation and industrial application For the industrial application of the adsorbents for the separation of light olefin/paraffin

mixtures various process options can be considered. Many of the current adsorption processes for other separations, like oxygen/nitrogen, are operated via pressure swing adsorption (PSA). An example of a flowsheet of a possible layout to obtain a high purity olefin is shown in Fig. 6.1 (Rege and Yang (2002); Ruthven, Farooq, and Knaebel (1994)). Starting from the lower blow down pressure at the end of step IV, the column is first pressurized with the binary mixture of the olefin and paraffin (I). Once the adsorption pressure is reached, the flow through the column is continued and the adsorbent is further loaded with the adsorptives. The paraffin will breakthrough and primarily the olefin will be adsorbed in the column (II). After a certain time the column is purged with a compressed olefin-rich stream (III). The purge stream removes the paraffin present in the void space between the adsorbent particles. Finally the olefin-rich product is obtained in the blow down step (IV).

The purge of the adsorption column (III) with an olefin rich stream will also effect the molar composition on the adsorbent. A displacement of adsorbed paraffin with the olefin will occur along the length of the column (Chapter 5 of this thesis). The higher affinity of the olefin results in the desorption of (part of) the paraffin. Therefore the olefin purge will result in an higher purity of the olefin product in the final step (IV) of the PSA-cycle (Tamura (1974)).

Compared to cryogenic distillation the energy cost will be lower (40-50% reduction in cryogenic distillation energy and 20-40% reduction in chiller/cooling duty of the feed (Eldridge (2005))). For the PSA cycle only energy costs will be required to pressurize the adsorption column and to compress part of the olefin-rich product in the purge-cycle (III). In cryogenic distillation, condensation of the entire feed mixture has to be achieved via multiple compression/expansion cycles. Furthermore recycle streams are present in the reboiler and condenser at the bottom and top of the cryogenic distillation tower (reflux-ratios up to 2.5-4 for ethylene/ethane and 12-20 for propylene/propane resulting in e.g. condenser duties >1 GJ/ton ethylene (Bouwmeester et al. (2005); Kroschwitz (1995); Reine (2004))).

132

Chapter 6

I II III IVPr

essu

rizat

ion

Adso

rptio

n

Co-

curr

ent p

urge

Cou

nter

-cur

rent

bl

ow d

own

Olefin/paraffin feed High purity olefin

Breakthrough product(Paraffin)

I II III IVPr

essu

rizat

ion

Adso

rptio

n

Co-

curr

ent p

urge

Cou

nter

-cur

rent

bl

ow d

own

Olefin/paraffin feed High purity olefin

Breakthrough product(Paraffin)

Fig. 6.1: Process flow sheet of the PSA cycles for the separation of olefin/paraffin mixtures (Rege and Yang (2002)).

The easiest PSA-cycles operate above atmospheric pressure, but lower blow down pressures can also be applied (e.g. Vacuum-PSA (Da Silva and Rodrigues (2001); Gomes and Hassan (2001))). For the highest working capacity of the adsorbent, the difference between the adsorbed amounts after the pressurisation (steps I-II) and blow down (step IV) should be large. For this criterion the more linear adsorption isotherm will yield the highest working capacity. Of the two adsorbents investigated in this study, the Faujasite zeolite without CuCl will then be more suited for such a separation process, since the large adsorption constant of the CuCl/Faujasite adsorbent results in a larger curvature of the isotherm.

Another important criterion for PSA operation will be the time required for the adsorption and the desorption process. Preferably these times are equal, so the PSA-cycles could be easily operated with only a limited number of adsorption columns. For the Faujasite zeolites steep profiles in the breakthrough profiles were observed and desorption was relatively fast even with our large Faujasite crystals (Chapter 5 of this thesis). Furthermore, binary adsorption equilibrium is reached quickly after the breakthrough of both components, reducing the possible losses in the adsorption capacity in the loading step (II).

The CuCl/Faujasite adsorbents require the operation at much lower operating pressures (e.g. Vacuum-PSA) or higher temperatures. At these conditions the isotherms are more linear and the adsorbent will be more selective compared to the other operating conditions and the blank Faujasite adsorbent. The operation at higher temperatures also reduces the diffusion limitations that were observed for the CuCl/Faujasite adsorbent. These limitations were more pronounced for propylene/propane mixtures and therefore a higher operating temperature will be required to achieve a fast diffusion rate and shorter contacting time. As a drawback, the

133


operation at higher temperatures will result in a lower adsorption capacity and therefore larger adsorption columns will be required.

Instead of the pressurization and blow down cycle shown in Fig. 6.1 other PSA-options are possible. Desorption can be achieved using an additional purge gas, e.g. nitrogen or C4-C5 hydrocarbons (Järvelin and Fair (1993); Thomas and Crittenden (1998)). A schematic flow sheet of such a process is shown in Fig. 6.2. First the column is loaded with the olefin/paraffin feed. The olefin will be adsorbed and the paraffin (and part of the purge gas still present in the column) will be obtained at the exit of the adsorption column. The mixture of the paraffin and the purge gas can then be separated or recycled as a diluted paraffin stream to for example a dehydrogenation reactor or cracker. Once the adsorption column is loaded with the olefin, counter-current desorption is performed using the purge gas. The olefin is then finally obtained after another separation step. Hereby the purge gas is recovered and can be reused in the separation process. This final separation of the olefin from the purge gas can be achieved using either an other adsorption process with a different adsorbent, membranes or via traditional distillation.

Adso

rptio

n

Olefin/paraffin feed

Paraffin + purge gas

Purge gas

Purge gas

Purge gas

Olefin + Purge gas

Olefin

Des

orpt

ion

Sepa

ratio

n

Sepa

ratio

n

Paraffin

Adso

rptio

n

Olefin/paraffin feed

Paraffin + purge gas

Purge gas

Purge gas

Purge gas

Olefin + Purge gas

Olefin

Des

orpt

ion

Sepa

ratio

n

Sepa

ratio

n

Paraffin

Fig. 6.2: Process flow sheet of a PSA operation with the aid of a purge gas (Thomas and Crittenden (1998)).

The adsorption and desorption can also be achieved by a change in the temperature of the adsorbent (Shu et al. (1990); Thomas and Crittenden (1998)). Adsorption is performed at low temperatures and the olefins are recovered at a higher temperature. As an advantage the desorption step of e.g. the CuCl/Faujasite adsorbent will be shorter and diffusion will be faster, in case of temperature swing adsorption (TSA). The temperature increase can be

134

Chapter 6

realized by heating the adsorption column or purging with a hot inert gas. In case the adsorption column is heated, small tubes will be required to allow a fast increase (or decrease) of the temperature of the adsorbent with the aid of external stream (or coolant). Furthermore, a good heat conducting adsorbent will be preferred.

For the application of the adsorbents in PSA and TSA operation the crystal size is also an important factor. The larger zeolite crystals used in this thesis resulted in a smaller pressure drop over the column. Commercial zeolite particles in adsorption column usually consist of smaller crystal combined to a larger particle. The smaller crystals will allow a shorter equilibrium with the internal pore structure of the zeolite. In the larger crystals, blockage of the pores could occur after the (incomplete) dispersion of CuCl (Chapter 5 of this thesis), and this will be reduced when smaller crystals are used.

Besides the cyclic PSA and TSA operations, future design options can also focus on the development of continuous processes. The zeolitic adsorbent can be synthesized as a membrane unit. The use of Faujasite zeolite membranes for the separation of propylene/propane mixtures has been investigated earlier by others (Giannakopoulos and Nikolakis (2005)). They showed that indeed an olefin/paraffin separation can be achieved over Faujasite membranes. For high selectivity a high quality membrane is required. Any pinhole in the membrane can seriously affect the separation performance. Diffusion through the pores of the zeolite is slower than the flow through these pinholes. Thinner and/or better membranes will allow a faster diffusion, but thinner membranes will also increase the chance of the formation of defects/pinholes. For membrane separation large surface areas will be required, since currently fluxes are still low. Research has been focussed on increasing these fluxes, while maintaining (or improving) the selectivity.

The dispersion of CuCl into a Faujasite membrane will result in the strong interaction of the olefins with CuCl via the -complex and therefore an increase of the selective adsorption of the olefins is expected. However, lower pressures on the permeate side of the membrane will be required to facilitate desorption. This can be achieved by applying a vacuum or using a large flow of a sweep gas. Alternatively, an operation at higher temperatures may be considered. This would result in a faster diffusion, but in a weaker adsorption.

Because of the slower diffusion in a CuCl/Faujasite membrane, the quality of the membrane should be higher than required for a pure Faujasite membrane. Otherwise the flow of both components through pinholes will dominate the total flux, resulting in a lower selectivity. For a good membrane, the smaller flux requires a larger membrane area to maintain the same production capacity as for the pure Faujasite membrane. Alternatively, research to Faujasite membranes with a larger flux can be considered. Perhaps CuCl dispersed membranes synthesized from other micro- or mesoporous materials, containing larger pores openings, is more advantageous than a CuCl/Faujasite membrane. Current research of the CuCl-dispersed materials is mainly limited to the cyclic PSA or TSA operations using adsorption columns.

During the adsorption tests performed in this thesis it was observed that oligomerization of the olefins towards carbon deposits can occur on the surface of the zeolite. These deposits are most likely formed on the few remaining acid sites of the zeolite. When theses sites are

135


quickly deactivated after the exposure to olefin, the long term exposure will probably not result in a continuing decrease of the performance of the adsorbent. Otherwise the long term operation in an industrial process can result in a decrease in the adsorption capacity and/or accessibility of the Faujasite zeolite. In membrane operation, the feed-to-zeolite ratio is relatively high compared to PSA and TSA processes and any detrimental effect due to these deposits will appear much earlier.

For the application of CuCl dispersed materials in industrial operation the stability of CuCl in the adsorbent will be an important criterion. As was shown in this research (Chapter 3 of this thesis) the exposure to ambient air resulted in the oxidation of the CuCl and the formation of fibres and cracks. Although the Cu+ adsorption site can be restored via a reduction step, the cracks and redistribution of the CuCl will be permanent. In steam cracking only traces of water can be expected to be present in the feed of the separation units. It was shown that water vapour, without oxygen, only affected the adsorption capacity of the adsorbent. After a long term PSA or membrane operation, the adsorption capacity may need to be restored by applying a heat treatment to the adsorbent. In case of a TSA operation the adsorbed water will automatically be desorbed during the higher temperature cycle. For membrane operation, the higher feed-to-zeolite ratio allows a shorter operation time before drying of the adsorbent will be needed. Operation of the membrane at elevated temperature, to create a higher flux and faster desorption, may also reduce the adverse effects of water adsorption.

The use of a continuous membrane separation will also allow a further process intensification of the olefin production plant. For the direct dehydrogenation of the paraffin, the dehydrogenation reactor can be combined with the membrane unit if both can operate at similar operation conditions. Due to the selective removal of the olefin from the reaction mixture, a shift of the conversion beyond the thermodynamic equilibrium can be achieved, reducing recycling streams and energy costs. In case CuCl dispersed membranes are used an oxydehydrogenation can not be combined with the membrane separation unit. The residual oxygen and produced water will react with CuCl and will lead to the destruction the zeolitic membrane, so a drying step will be required before the separation can be performed in a separate unit.

For the production of olefins via e.g. steam cracking the entire separation scheme (Chapter 1 of this thesis) could also be completely rearranged. After the removal of the acidic gases (CO2 and H2S) and drying, acetylene has to be removed by selective catalytic hydrogenation. Acetylene and ethylene form an azeotrope in cryogenic distillation and would limit the purity of ethylene. Furthermore, acetylene and dienes have a much stronger tendency to polymerize on the adsorbent. Further, in case CuCl dispersed adsorbents would be used, the desorption time will be severely hindered by the much stronger adsorption of acetylene on CuCl (Hübner et al. (2003)). After this selective hydrogenation a split between the unsaturated and saturated hydrocarbons & hydrogen can be achieved. The saturated hydrocarbons (methane, ethane, propane and butane) & hydrogen can then be separated via conventional cryogenic distillation or perhaps alternative adsorption processes. The unsaturated hydrocarbons (ethylene,

136

Chapter 6

propylene and C4-olefins) are adsorbed on the Faujasite or CuCl/Faujasite adsorbent. After their recovery, they can be split via conventional cryogenic distillation or an alternative adsorption process. Since the relative volatilities will differ much more, a cryogenic separation of these two mixtures will be less complex and can occur over shorter distillation columns containing fewer separation trays.

6.3 Conclusions In this study we investigated the development, optimization and application of Faujasite

and CuCl modified Faujasite zeolites. The current literature about the dispersion of CuCl in Faujasite zeolites was limited to only a few papers, which mainly focused on the basic preparation route. In chapters 2 and 3 of this thesis a more thorough investigation of the various factors affecting the preparation and stability of the CuCl/Faujasite adsorbents was performed.

In the literature no direct comparison between Faujasite and CuCl/Faujasite adsorbent for the adsorptive separation of ethylene/ethane or propylene/propane mixtures was performed. Furthermore, adsorption data for Faujasite zeolites is often limited to single component data and binary adsorption data of these olefin/paraffin mixtures is limited to a few data points. With this research more knowledge about both the single and binary adsorption on Faujasite and CuCl/Faujasite adsorbents is obtained (Chapters 4 and 5 of this thesis). The IAS-theory describes the binary mixture adsorption on the NaX zeolite well, while quantitative deviations from the ideal predictions were observed for the CuCl/Faujasite adsorbent. It nevertheless forms a good basis for the design of adsorptive separation processes, either for discontinuous PSA or for continuous membrane operations.

Both adsorbents have their advantages and disadvantages. The CuCl/Faujasite adsorbent has a stronger affinity for the olefin. This results in a higher selectivity, but it also requires the operation at higher temperatures or sub-atmospheric pressures to be able to desorb the olefin, since the olefin is the desired product of the separation process. At these conditions the loading in the CuCl/Faujasite will be lower and therefore larger separation units than at ambient conditions will be required. For the application in an industrial process the stability of the CuCl/Faujasite in ambient atmospheres will affect the handling of the material during the filling of the separation unit or will require the presence of an inert atmosphere during temporary shutdowns or maintenance.

The Faujasite adsorbent has a lower affinity, and therefore lower selectivity for the olefin. Therefore more separation steps will be needed to obtain the required purity. The advantage of this adsorbent is that the desorption of the olefin is much easier. Furthermore, the diffusion of the components in and out of the zeolite pores was much faster compared to the CuCl/Faujasite adsorbent. The slower diffusion of the CuCl/Faujasite adsorbent may be reduced by the use of a Cu+ ion-exchange Faujasite zeolite (Takahashi et al. (2001)), but the other disadvantages of the Cu+/Faujasite adsorbent will still remain.

For these reasons Faujasite would be the preferred choice to allow a fast and easy separation process based on adsorption. Application of CuCl/Faujasite as adsorbent would be

137


hampered by the too strong affinity for the olefin, the slower diffusion of all components and its stability.

6.4 References Bouwmeester, P., Borg, N., Van der Knaap, J., Vermeer, B., and De Wit, M., CPD-report

3317: Design of a Process to Manufacture Ethene from Ethane by means of Oxidative Dehydrogenation, DelftChemTech, Delft University of Technology (2005).

Da Silva, F. A. and Rodrigues, A. E., Propylene/Propane Separation by Vacuum Swing Adsorption using 13X Zeolite, AIChE J. 47 (2001) 341-357.

Eldridge, R. B., Brainstorming Session Background Information as Part of the Hybrid Technology Workshop Separations Research Program in Austin, TX, USA, (2005).

Giannakopoulos, I. G. and Nikolakis, V., Separation of Propylene/Propane Mixtures using Faujasite-Type Zeolite Membranes, Ind. Eng. Chem. Res. 44 (2005) 226-230.

Gomes, V. G. and Hassan, M. M., Coalseam Methane Recovery by Vacuum Swing Adsorption, Sep. Purif. Technol. 24 (2001) 189-196.

Hübner, G., Rauhut, G., Stoll, H. and Roduner, E., Ethyne Adsorbed on CuNaY Zeolite: FTIR Spectra and Quantum Chemical Calculations, J. Phys. Chem. B 107 (2003) 8568-8573.

Järvelin, H. and Fair, J. R., Adsorptive Separation of Propylene-Propane Mixtures, Ind. Eng. Chem. Res. 32 (1993) 2201-2207.

Kroschwitz, J. I., Kirk-Othmer Encyclopedia of Chemical Technology, Wiley, New York (1995) Vol. 9 pp. 898.

Rege, S. U. and Yang, R. T., Propane/Propylene Separation by Pressure Swing Adsorption: Sorbent Comparison and Multiplicity of Cyclic Steady States, Chem. Eng. Sci. 57 (2002) 1139-1149.

Reine, T.A., Olefin/Paraffin Separation by Reactive Adsorption, PhD Thesis, The University of Texas, Austin, USA (2004).


Shu, C. M., Kulvaranon, S., Findley, M. E. and Liapis, A. I., Experimental and Computational Studies on Propane-Propylene Separation by Adsorption and Variable-Temperature Stepwise Desorption, Separ. Technol. 1 (1990) 18-28.

Takahashi, A., Yang, R. T., Munson, C. L. and Chinn, D., Cu(I)-Y-Zeolite as a Superior Adsorbent for Diene/Olefin Separation, Langmuir 17 (2001) 8405-8414.

Tamura, T., Adsorption Process for Gas Separation, US Patent 3 797 201 (1974). Thomas, W. J. and Crittenden, B., Adsorption Technology and Design, Butterworth-

Heinemann, Oxford (1998).

138

Samenvatting

Adsorptieve Scheiding

van

Laagmoleculaire Olefine/Paraffine Mengsels

Dispersie van CuCl in Faujasite Zeolieten

Samenvatting

Laagmoleculaire onverzadigde koolwaterstoffen, zoals etheen en propeen zijn belangrijke grondstoffen voor de chemische industrie. Ze worden verkregen via thermische kraakprocessen gecombineerd met energie-intensieve cryogene destillatie van de geproduceerde olefine/paraffine mengsels. Als minder energie-intensief alternatief, kan deze scheiding worden gerealiseerd via een selectief adsorptieproces. Daarvoor is het van belang om een goedkoop en effectief adsorbens te vinden.

Het centrale thema van dit proefschrift is de ontwikkeling en optimalisatie van een adsorbens voor de selectieve scheiding van olefine/paraffine mengsels en het testen van de toepassing van deze materialen. Hiervoor werd een op zeolieten gebaseerd adsorbens gebruikt (Faujasiet NaX of NaY). Er werden grote NaX zeolietkristallen gesynthetiseerd (Hoofdstuk 2 van dit proefschrift) om zo bij toepassing in de doorbraak-opstelling een lagere drukval over de gepakte kolom te hebben. Het gebruik van grote kristallen resulteert ook in een vermindering van het aantal model-variabelen die het scheidingsgedrag beïnvloeden, omdat dit in tabletten opgebouwd uit kleinere kristallen, zal worden beïnvloed door de additionele transportverschijnselen binnenin het tablet.

In de zeolietkristallen werd succesvol CuCl gedispergeerd. Er werden verzadigingscapaciteiten van 36 gew% CuCl voor NaX en 43 gew% voor NaY gevonden, wat overeenkomt met ongeveer 10 CuCl moleculen per ‘super-cavity’.

FTIR spectroscopie met CO als ‘probe’ molecuul (Hoofdstuk 2 van dit proefschrift) bevestigde dat er inderdaad een dispersie van CuCl in het zeoliet plaatsvindt. Tevens werd een sterke adsorptie van CO op CuCl via -complexatie waargenomen. Adsorptie van ethaan, etheen, propaan en propeen in de lage druk IR transmissie-cel bevestigde dat de adsorptie op NaY zeoliet zwak is voor alle componenten. Door een drukverlaging kunnen de

139

Samenvatting

geadsorbeerde gassen snel worden verwijderd van het Faujasiet zeoliet. Op de CuCl/Faujasiet adsorbentia laten de olefines een veel sterkere adsorptie zien. Er werden additionele absorptie-banden gezien met IR spectroscopie, overeenkomend met het -complex van Cu+ met de dubbele bindingen van de olefines. Ten opzichte van het Faujasiet zeoliet vond de complete desorptie van de olefines van dit complex plaats bij veel lagere drukken en/of hogere temperaturen.

Naast het potentieel van het adsorbens om olefine/paraffine mengsels te scheiden, is het ook van belang dat het materiaal stabiel blijft voor langere tijd. In industriële voedingen kunnen zich sporen van waterdamp in het gasmengsel bevinden. Er werd aangetoond dat de blootstelling van het CuCl/Faujasiet adsorbens aan (vochtige) lucht resulteert in de geleidelijke vernietiging van het adsorbens. CuCl werd geoxideerd tot CuCl2·3Cu(OH)2

(Atacamite) toen het werd blootgesteld aan een mengsel van waterdamp en zuurstof (Hoofdstuk 3 van dit proefschrift). Er vindt een expansie plaats van het Cu-materiaal binnenin de poriestructuur van het zeoliet door de lagere molaire dichtheid (mol Cu per m3) van het geoxideerde Cu-materiaal. De restrictie binnenin de zeolietporiën resulteerde in de groei van draden uit de buitenste poriën van het zeolietkristal en/of in de vorming van scheuren in het zeolietkristal. Een grotere hoeveelheid CuCl in de poriën en holtes van het zeoliet resulteerde in een langere stabiliteit van het adsorbens. Dit kan worden verklaard door de verminderde diffusiesnelheid van water en zuurstof in het zeoliet door de aanwezigheid van CuCl. Boven de verzadigingscapaciteit zal er nog CuCl aanwezig zijn buiten het zeolietkristal, dat snel zal reageren met water en zuurstof. Daarom zou voor de maximale stabiliteit de belading van CuCl dichtbij, maar niet boven de verzadigingscapaciteit moeten zijn.

De aanwezigheid van waterdamp, zonder de aanwezigheid van zuurstof, leidt tot een vermindering van de adsorptiecapaciteit van CO en andere gassen op het CuCl/Faujasiet. Het water bedekt de adsorptieplaatsen in het adsorbens en/of limiteert de toegang van de adsorberende componenten tot de zeolietporiën. Na de verwijdering van het geadsorbeerde water door een warmtebehandeling was de snelle adsorptie van CO hersteld. Voor de industriële toepassing zou de blootstelling aan lucht vermeden moeten worden, terwijl alleen de aanwezigheid van sporen van waterdamp niet noodzakelijkerwijs destructief is voor het CuCl/Faujasiet adsorbens.

Één-component adsorptie-isothermen toonden aan dat zowel NaX als CuCl/NaX een grotere affiniteit voor de olefines hebben (Hoofdstuk 4 van dit proefschrift). De dispersie van CuCl in het Faujasiet resulteerde in een vermindering van de adsorptiecapaciteit voor zowel de olefines als de paraffines. Dit kan worden verklaard door de reductie van het beschikbare porievolume wanneer CuCl wordt gedispergeerd op de wanden van de zeolietporiën. Bij lagere drukken werden voor het CuCl/Faujasiet adsorbens hogere (of minder gereduceerde) beladingen van de olefines gevonden dan voor het pure Faujasiet adsorbens, doordat de olefines, via een sterke -complexatie, een hogere affiniteit hebben voor CuCl. Er werd een hogere adsorptieconstante en adsorptiewarmte gevonden voor de olefines op CuCl/NaX door deze interactie. Ongeveer 17% van het Cu+ is betrokken bij deze -complexatie. De metingen van de adsorptie-isothermen via de volumetrische methode lieten tevens zien dat het bereiken

140

Samenvatting

van evenwicht langer duurde nadat CuCl in de poriën van NaX is gedispergeerd, wat indicatief is voor een langzamere diffusie van de adsorberende componenten in het adsorbens.

De selectieve adsorptie van de olefines werd ook bevestigd in de doorbraak-opstelling. Er werden binaire adsorptie-data bepaald met deze opstelling tussen 0.8-54 kPa en bij 318, 358 en 408 K (Hoofdstuk 5 van dit proefschrift). Op het CuCl/NaX adsorbens werd de adsorptie van de koolwaterstoffen beïnvloed door diffusielimitaties. Beneden 358 K zijn deze diffusie-limitaties duidelijker aanwezig, met name voor de relatief grote propaan en propeen moleculen. Bij hogere temperaturen zijn de diffusielimitaties verminderd, maar worden tevens de adsorptiebeladingen van het CuCl/NaX adsorbens kleiner. De hoogste selectiviteiten werden verkregen bij lage (sub-atmosferische) partiële drukken. Voor etheen resulteert de dispersie van CuCl in de poriën van het NaX zeoliet in een veel grotere toename van de selectiviteit dan voor propeen. De mengsel-selectiviteit voor etheen nam toe van 3-10 op NaX tot 15-200 op CuCl/NaX, terwijl de mengsel-selectiviteit voor propeen toenam van 3-7 op NaX tot 8-30 op CuCl/NaX. Het kleinere etheen molecuul diffundeert veel makkelijker door de smalle poriën naar het centrum van het grote zeolietkristal. Bovendien, de interactie van etheen met CuCl via -complexatie is veel sterker dan voor propeen, omdat propeen een extra methyl groep heeft. Wanneer de selectiviteit onafhankelijk is van de kristalgrootte, dan kunnen, om een sneller industrieel proces mogelijk te maken, de diffusielimitaties worden verminderd door kleine kristallen te gebruiken. Om toch een lage drukval over de adsorptiekolom te behouden, kunnen deze kleinere kristallen worden samengeperst tot grote deeltjes.

Op NaX, zonder CuCl, werd ook een selectieve adsorptie van de olefines waargenomen. Het binaire evenwicht werd relatief snel bereikt na de doorbraak van beide componenten, wat suggereert dat de diffusie veel sneller is in het NaX zeoliet. De lagere selectiviteit voor NaX blijft bijna constant binnen het onderzochte drukinterval (0.8-54 kPa). Voor een binair (50:50) mengsel van propeen en propaan werd een overgang waargenomen van een enthalpie-gecontroleerde adsorptie bij lage beladingen naar een adsorptie die beïnvloed wordt door een pakkingsentropie effect bij hogere beladingen, met een gunstig effect op de selectiviteit. Een zelfde overgang en positief effect op de selectiviteit werd ook waargenomen voor etheen/ethaan mengsels op 36 gew% CuCl/NaX. Dit is een van de eerste keren dat zo’n invloed experimenteel is waargenomen.

De mengsel adsorptie van het binaire (50:50) olefine/paraffine mengsel op NaX kan goed worden beschreven met de ‘Ideal Adsorbed Solution’ (IAS) theorie (Hoofdstuk 1 en 5 van dit proefschrift) en de enkel component isothermen, zowel kwalitatief als kwantitatief. Voor het 36 gew% CuCl/NaX adsorbens werden grotere verschillen met de voorspellingen van de IAS-theorie gevonden, waarschijnlijk door de aanwezigheid van diffusielimitaties, of door de niet-ideale adsorptie via -complexatie; niettemin komen de kwalitatieve trends goed overeen. De waargenomen overgang van een enthalpie gecontroleerde naar een entropie-beïnvloedeadsorptie wordt inderdaad voorspeld met de IAS-theorie.

Op basis van deze resultaten werden de mogelijkheden en problemen voor industriële toepassing van de adsorbentia in enkele procesopties, waaronder PSA, TSA en membraan-toepassingen, bediscussieerd (Hoofdstuk 6 van dit proefschrift). Van de twee adsorbentia die

141

Samenvatting

in dit proefschrift werden onderzocht, liet het CuCl/Faujasiet adsorbens een hogere selectiviteit zien ten opzichte van het Faujasiet-adsorbens, wat een gunstig effect heeft op het aantal scheidingsstappen. Door de sterkere adsorptie zou het gebruik van CuCl/Faujasiet de operatie bij hogere temperatuur of sub-atmosferische druk vereisen. Maar bij deze condities zal de adsorptiecapaciteit lager zijn. Bovendien wordt de adsorptie beïnvloed door diffusielimitaties. Deze nadelen maken de toepassing van CuCl/Faujasiet in processen meer complex en minder aantrekkelijk. Daarom wordt het pure Faujasiet gezien als een beter materiaal en dit zou bijvoorbeeld gebruikt kunnen worden in PSA of membraan-toepassingen voor het scheiden van mengsels van etheen/ethaan of propeen/propaan.

142

List of publications and presentations


Published papers

Van Miltenburg, A, Poot, W, De Loos, TW, High-Pressure phase behavior of the binary system of isobutane and diamantane, Journal of Chemical Engineering Data, 45 (2000), 977-979.

Zhu, W, Groen, JC, Van Miltenburg, A, Kapteijn, F, Moulijn, JA, Comparison of adsorption behaviour of light alkanes and alkenes on Kureha activated carbon, Carbon, 43 (2005), 1416-1423.

Van Miltenburg, A, Zhu, W, Kapteijn, F, Moulijn, JA, Zeolite based separation of light olefin and paraffin mixtures, Studies in Surface Science and Catalysis, 158B (2005), 979-986.

Van Miltenburg, A, Zhu, W, Kapteijn, F, Moulijn, JA , Adsorptive separation of light olefin/paraffin mixtures, Chemical Engineering Research and Design 84 (A5) (2006), 350-354.

Paper submitted

Van Miltenburg, A, Gascon, J, Zhu, W, Kapteijn, F, Moulijn, JA, Propylene/propane mixture adsorption on Faujasite adsorbents, Adsorption, submitted.

Papers in preparation

The research presented in the chapter 2 to 5 will be submitted in further publications.

143


Oral presentations

Van Miltenburg, A, Zhu, W, Kapteijn, F, Moulijn, JA, Synthesis of zeolite-based selective sorbents for alkane/alkene separation, Workshop of the Dutch Zeolite Association, Delft, The Netherlands, December 2003.

Van Miltenburg, A, Zhu, W, Kapteijn, F, Moulijn, JA, Adsorptive separation of light olefins and paraffins, 8th Fundamentals Of Adsorption Congress (FOA8), Sedona (AZ), United States of America, May 2004.

Van Miltenburg, A, Zhu, W, Kapteijn, F, Moulijn, JA, Separation of light olefins and paraffins, 4th Netherlands Process technology Symposium (NPS4), Veldhoven, The Netherlands, October 2004.

Van Miltenburg, A, Zhu, W, Kapteijn, F, Moulijn, JA, Separation of light olefin/paraffin mixtures, Adsorption Course 2005 / 2nd workshop of the International Research Training Group (IRTG), Eindhoven, The Netherlands, April 2005.

Van Miltenburg, A, Zhu, W, Kapteijn, F, Moulijn, JA, Zeolite based separation of light olefin and paraffin mixtures, 3rd International Conference of the Federation of European Zeolite Associations (FEZA3), Prague, Czech Republic, August 2005.

Van Miltenburg, A, Zhu, W, Kapteijn, F, Moulijn, JA, Analysis of adsorption & diffusion in zeolite A with the MultiTrack, 3rd workshop of the International Research Training Group (IRTG), Leipzig, Germany, September 2005.

Heibel, AK, Du, P, Van Miltenburg, A, Kapteijn, F, Moulijn, JA, External mass transfer phenomena in square channel monolith structures under film-flow conditions, 2nd

International Conference on Structured Catalysts and Reactors (ICOSCAR-2), Delft, The Netherlands, October 2005.

Gascon, J, Van Miltenburg, A, Zhu, W, Kapteijn, F, Moulijn, JA, Separation of propylene/propane mixtures over microporous materials, 8th Netherlands Catalysis & Chemistry Conference (NCCC-VIII), Noordwijkerhout, The Netherlands, March 2007.

144


Poster presentations

De Lathouder, KM, Van Miltenburg, A, Kapteijn, F, Moulijn, JA, Biocatalytic liquid phase hydrolysis in a monolith reactor, 3rd Netherlands Catalysis & Chemistry Conference (NCCC-III), Noordwijkerhout, The Netherlands, March 2002.

Van Miltenburg, A, Zhu, W, Kapteijn, F, Moulijn, JA, Shape Selective light olefin/paraffin separation, 4th Netherlands Catalysis & Chemistry Conference (NCCC-IV), Noordwijkerhout, The Netherlands, March 2003.

Van Miltenburg, A, Zhu, W, Kapteijn, F, Moulijn, JA, Shape Selective light olefin/paraffin separation, NATO Advanced Study Institute: Fluid Transport in Nanoporous Materials, La Colle-sur-Loup (Nice), France, June 2003.

Van Miltenburg, A, Zhu, W, Kapteijn, F, Moulijn, JA, Synthesis of zeolite-based selective adsorbents for the separation of light olefin/paraffin mixtures, 14th International Zeolite Conference (IZC14), Cape Town, South Africa, April 2004.

Van Miltenburg, A, Zhu, W, Kapteijn, F, Moulijn, JA, Modified zeolites for olefin / paraffin separation, 6th Netherlands Catalysis & Chemistry Conference (NCCC-VI), Noordwijkerhout, The Netherlands, March 2005.

Van Miltenburg, A, Zhu, W, Kapteijn, F, Moulijn, JA, Adsorptive separation of light olefin/paraffin mixtures, Sustainable (Bio)Chemical Process Technology, Incorporating the 6th International Conference on Process Intensification, Delft, The Netherlands, September 2005.

Gascon, J, Van Miltenburg, A, Zhu, W, Kapteijn, Moulijn, JA, Propylene/propane mixture adsorption on Faujasite, 6th Netherlands Process technology Symposium (NPS6), Veldhoven, The Netherlands, October 2006.

Gascon, J, Van Miltenburg, A, Zhu, W, Kapteijn, Moulijn, JA, Propylene/propane mixture adsorption on Faujasite, 2nd International School and Workshop on IN-Situ Study and Development of Processes Involving Porous Solids (NoE INSIDE-PORES), Thessaloniki, Greece, February 2007.

Van Miltenburg, A, Gascon, J, Zhu, W, Kapteijn, F, Moulijn, JA, Light olefin/paraffin mixture adsorption on Faujasite adsorbents, 9th Fundamentals Of Adsorption Congress (FOA9), Giardini Naxos (Sicily), Italy, May 2007.

145


Van Miltenburg, A, Blom, W., Gascon, J, Ferreira, A.F.P., Zhu, W, Kapteijn, F, Moulijn, JA, Light olefin/paraffin mixture adsorption on Faujasite sorbents, International Symposium on Catalysis Engineering, Delft, The Netherlands, June 2007.

146

Dankwoord

Dankwoord

Hierbij wil ik graag iedereen bedanken die heeft meegeholpen aan de totstandkoming van dit proefschrift. In eerste instantie natuurlijk mijn promotoren Freek Kapteijn en Jacob Moulijn. Freek die mij de mogelijkheid bood voor een promotieplaats, zijn kritische kijk op dit proefschrift en voor zijn internationale onderzoekscontacten, zoals de IRTG “Diffusion in Porous Materials” en NoE “Inside Pores”. Jacob om zijn laatste kritische kijk en zijn voorliefde voor ‘mooie plaatjes’. Weidong thanks for your input, ideas and critical comments concerning this research and/or manuscripts.

Koos Jansen wil ik bedanken voor de lange discussies die we hebben gehad over de syntheses van DD3R en Faujasiet. Zelfs met een overvolle agenda wist je altijd wel een gaatje te vinden en toonde je veel belangstelling als je weer eens een rondje door het ‘Pore’-lab liep.

Verder wil ik de werkplaats, en dan met name Rien Slooter, bedanken voor hun assistentie tijdens de bouw van de doorbraakopstelling. Uiteindelijk is het een mooie opstelling geworden. Het was wel een hele puzzel om alles lekdicht te krijgen, zeker omdat er soms een deel van de ferrule ontbrak.

Ook wil ik de studenten bedanken die hebben deelgenomen aan verschillende facetten van mijn onderzoek. Arnoud Greidanus dank voor de vele syntheses van NaX. Mede dankzij jouw onderzoek kon ik de grote en uniforme zeolietkristallen maken die beschreven staan in dit proefschrift. Ik wens je net zoveel succes met de rest van je studie als met het roeien. Yan (Irena) Jiao, thank you for your preliminary IR-research on the adsorption of olefins and paraffins. Your work allowed me to perform part of the IR-results presented in this thesis in just a couple of weeks. It is a pity that personal reasons made you decide to stop your studies, so close to the finish. John Juan bedankt voor jouw onderzoek naar de synthese van

147

Dankwoord

Chabaziet. Helaas lukte het je niet om het gewenste zeoliet te maken. Ik wens je veel succes met het laatste deel van je studie.

Verder wil ik natuurlijk mijn kamergenoten bedanken voor de vele ‘wetenschappelijke’ discussies. Allereerst Peng Du met wie ik toch de langste tijd de kamer heb gedeeld. Volgens mij hebben wij ‘ons’ kantoor in de afgelopen jaren wel tig keer verbouwd en opnieuw ingericht. Hirokazu Shibata thanks for introducing me into parts of your electrochemistry research during the various discussions, the nice dinners, for your assistance as paranimf and for bringing the NIOK soccer cup to Delft. It’s a pity the results of the R&CE teams were not so good in Delft the next year. Finally I like to thank Joana Carneiro. We shared the office for only a short period at the end of my stay in Delft. I whish you success with your research and I guess Arnoud is a good student for starting your research.

Bart van der Linden en Harrie Jansma wil ik bedanken voor de technische ondersteuning. Bart in het bijzonder voor zijn assistentie en expertise met IR spectroscopie en de vele keren bij het uit elkaar halen en in elkaar zetten van de microGC. I also like to thank Jorge Gascon for checking the few remaining questions with some of the IR-experiments.

Sander Brouwer en Johan Groen wil ik bedanken voor hun assistentie met de volumetrische adsorptie metingen. Johan wil ik ook bedanken voor zijn input en discussies tijdens ons vervolgonderzoek naar de vorming van koperdraden uit zeolieten.

Patricia Kooyman wil ik bedanken voor de TEM fotos. Gezien de hoge mobiliteit van het CuCl in de zeolieten was het lastig om een foto te maken waarop geen draadvorming of een explosie van koperdeeltjes te zien was.

Tevens bedank ik hierbij de secretaresses Els Arkesteijn en Elly Hilkhuijsen, die steeds de gaatjes in beide overvolle agenda’s van Freek en Jacob wisten te combineren. Elly wil ik ook bedanken voor de laatste afwikkeling van de promotie formulieren.

Verder wens ik de overige collega’s veel succes. Jullie hielden me af en toe goed van het werk. Bijvoorbeeld door steeds op de deur te kloppen met vragen of problemen met betrekking tot elektronica, Labview, leidingwerk of gewoon voor een wetenschappelijke discussie met een ‘Top-NIOK-doctor’. Het was wel leuk en leerzaam, maar soms kon ik maar beter thuis aan dit proefschrift werken.

I like to thank my new colleagues at SINTEF, who allowed me to spend time on the completion of this thesis, for their patience. Takk skal dere ha!

Tenslotte wil ik mijn ouders en broer bedanken voor hun interesse, input, geduld en de opbeurende woorden. Neef Evthimios Karaliolios wil ik bedanken voor zijn assistentie als paranimf. Ook bedank ik oma en de ooms, tantes en overige neefjes en nichtjes, die ik iedere keer weer, zonder gebruik van vakjargon, moest vertellen dat ik nu met zeolieten bezig was en niet, zoals tijdens mijn afstuderen, met monolieten. Er is nogal een groot verschil in dimensies tussen deze twee materialen. Jullie weten reeds wat monolieten zijn en hopelijk schept dit proefschift meer duidelijkheid over zeolieten.

Arjen van Miltenburg Oslo, juni 2007

148

Curriculum Vitae

Curriculum Vitae

Arjen van Miltenburg werd geboren op 1 oktober 1977 te Utrecht. In 1996 behaalde hij zijn VWO diploma aan het Krimpenerwaard College in Krimpen aan den IJssel. In datzelfde jaar begon hij met de studie Scheikundige Technologie aan de Technische Universiteit Delft. In 2001 studeerde hij cum laude af op de vloeistof-vast stofoverdracht bij een enzymatische conversie in een monoliet reactor binnen de sectie Industrial Catalysis onder begeleiding van diplom. ir. Achim Heibel, prof. dr. Freek Kapteijn en prof. dr. Jacob Moulijn. In datzelfde jaar begon hij zijn promotieonderzoek in die sectie, ondertussen uitgebreid tot Reactor and Catalysis Engineering, onder begeleiding van prof. dr. Freek Kapteijn, prof. dr. Jacob Moulijn en dr. Weidong Zhu. De resultaten van dit onderzoek staan beschreven in dit proefschrift.

Sinds 1 oktober 2006 is hij werkzaam als postdoc bij SINTEF in Oslo, Noorwegen. Hij doet daar onderzoek naar de modificatie en toepassing van MCM-22 zeolieten binnen het Europese Marie-Curie onderzoeksprogramma “INtelligent DEsign of Nanoporous Sorbents (INDENS)”.

149

150

Date post:	11-Sep-2021
Category:	Documents
Upload:	others
View:	3 times
Download:	0 times

Adsorptive Separation of Light Olefin/Paraffin Mixture. Dispersion of CuCl in Faujasite Zeolites

Documents